EP3665648A2 - Détection de mise en place et enlèvement d'article à l'aide d'une reconnaissance d'image - Google Patents

Détection de mise en place et enlèvement d'article à l'aide d'une reconnaissance d'image

Info

Publication number
EP3665648A2
EP3665648A2 EP18844384.0A EP18844384A EP3665648A2 EP 3665648 A2 EP3665648 A2 EP 3665648A2 EP 18844384 A EP18844384 A EP 18844384A EP 3665648 A2 EP3665648 A2 EP 3665648A2
Authority
EP
European Patent Office
Prior art keywords
images
cameras
identified
subjects
sequences
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP18844384.0A
Other languages
German (de)
English (en)
Other versions
EP3665648A4 (fr
Inventor
Jordan E. Fisher
Daniel L. Fischetti
Brandon L. Ogle
John F. Novak
Kyle E. Dorman
Kenneth S. Kihara
Juan C. Lasheras
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Standard Cognition Corp
Original Assignee
Standard Cognition Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US15/847,796 external-priority patent/US10055853B1/en
Priority claimed from US15/907,112 external-priority patent/US10133933B1/en
Priority claimed from US15/945,466 external-priority patent/US10127438B1/en
Priority claimed from US15/945,473 external-priority patent/US10474988B2/en
Application filed by Standard Cognition Corp filed Critical Standard Cognition Corp
Publication of EP3665648A2 publication Critical patent/EP3665648A2/fr
Publication of EP3665648A4 publication Critical patent/EP3665648A4/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q20/00—Payment architectures, schemes or protocols
    • G06Q20/08—Payment architectures
    • G06Q20/20—Point-of-sale [POS] network systems
    • G06Q20/203—Inventory monitoring
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q20/00—Payment architectures, schemes or protocols
    • G06Q20/08—Payment architectures
    • G06Q20/20—Point-of-sale [POS] network systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q20/00—Payment architectures, schemes or protocols
    • G06Q20/08—Payment architectures
    • G06Q20/20—Point-of-sale [POS] network systems
    • G06Q20/208—Input by product or record sensing, e.g. weighing or scanner processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G07—CHECKING-DEVICES
    • G07G—REGISTERING THE RECEIPT OF CASH, VALUABLES, OR TOKENS
    • G07G1/00—Cash registers
    • G07G1/0036—Checkout procedures
    • G07G1/0045—Checkout procedures with a code reader for reading of an identifying code of the article to be registered, e.g. barcode reader or radio-frequency identity [RFID] reader
    • G07G1/0054—Checkout procedures with a code reader for reading of an identifying code of the article to be registered, e.g. barcode reader or radio-frequency identity [RFID] reader with control of supplementary check-parameters, e.g. weight or number of articles
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00—Diagnosis, testing or measuring for television systems or their details
    • H04N17/002—Diagnosis, testing or measuring for television systems or their details for television cameras
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/90—Arrangement of cameras or camera modules, e.g. multiple cameras in TV studios or sports stadiums

Definitions

  • the present invention relates to systems and components thereof usable for cashier-less checkout.
  • a difficult problem in image processing arises when images from multiple cameras disposed over large spaces are used to identify and track actions of subjects.
  • a system, and method for operating a system are provided for tracking changes by subjects, such as persons, in an area of real space, and other complex interactions of subjects with their environments, using image processing.
  • This function of tracking changes by image processing presents a complex problem of computer engineering, relating to the type of image data to be processed, what processing of the image data to perform, and how to determine actions from the image data with high reliability.
  • the system described herein can perform these functions using only images from cameras disposed overhead in the real space, so that no retrofitting of store shelves and floor space with sensors and the like is required for deployment in a given setting.
  • a system and method are provided for tracking put and takes of inventory items by subjects in an area of real space including inventory display structures that comprise using a plurality of cameras disposed above the inventory display structures to produce respective sequences of images of inventory display structures in corresponding fields of view in the real space, the field of view of each camera overlapping with the field of view of at least one other camera in the plurality of cameras.
  • a system and method are described for detecting puts and takes of inventory items by identifying semantically significant changes in the sequences of images relating to inventory items on inventory display structures and associating the semantically significant changes with subjects represented in the sequences of images.
  • a system and method are provided for tracking put and takes of inventory items by subjects in an area of real space, that comprise using a plurality of cameras disposed above the inventory display structures to produce respective sequences of images of inventory display structures in corresponding fields of view in the real space, the field of view of each camera overlapping with the field of view of at least one other camera in the plurality of cameras.
  • a system and method are described for detecting puts and takes of inventory items by identifying gestures of subjects and inventory items associated with the gestures by processing foreground data in the sequences of images.
  • a system and method that combines foreground processing and background processing in the same sequences of images.
  • the system and method provided include using these sequences of images for detecting puts and takes of inventory items by identifying gestures of subjects and inventory items associated with the gestures by processing foreground data in the sequences of images; and using these sequences of images for detecting puts and takes of inventory items by identifying semantically significant changes in the sequences of images relating to inventory items on inventory display structures by processing background data in the sequences of images, and associating the semantically significant changes with subjects represented in the sequences of images.
  • the system uses a plurality of cameras to produce respective sequences of images of corresponding fields of view in the real space.
  • the field of view of each camera overlaps with the field of view of at least one other camera in the plurality of cameras.
  • the system includes first image processors, including subject image recognition engines, receiving corresponding sequences of images from the plurality of cameras.
  • the first images processors process images to identify subjects represented in the images in the corresponding sequences of images.
  • the system further includes, second image processors, including background image recognition engines, receiving corresponding sequences of images from the plurality of cameras.
  • the second image processors mask the identified subjects to generate masked images, and process the masked images to identify and classify background changes represented in the images in the corresponding sequences of images.
  • the background image recognition engines comprise convolutional neural networks.
  • the system includes logic to associate identified background changes with identified subjects.
  • the second image processors include a background image store to store background images for corresponding sequences of images.
  • the second image processors further include mask logic to process images in the sequences of images to replace foreground image data representing the identified subjects with background image data.
  • the background image data is collected from the background images for the corresponding sequences of images to provide the masked images.
  • the mask logic combines sets of N masked images in the sequences of images to generate sequences of factored images for each camera.
  • the second image processors identify and classify background changes by processing the sequence of factored images.
  • the second image processors include logic to produce change data structures for the corresponding sequences of images.
  • the change data structures include coordinates in the masked images of identified background changes, identifiers of an inventory item subject of the identified background changes and classifications of the identified background changes.
  • the second image processors further include coordination logic to process change data structures from sets of cameras having overlapping fields of view to locate the identified background changes in real space.
  • the classifications of identified background changes in the change data structures indicate whether the identified inventory item has been added or removed relative to the background image.
  • the classifications of identified background changes in the change data structures indicate whether the identified inventory item has been added or removed relative to the background image.
  • the system further includes logic to associate background changes with identified subjects.
  • the system includes the logic to make detections of takes of inventory items by the identified subj ects and of puts of inventory items on inventory display structures by the identified subjects.
  • the system includes logic to associate background changes with identified subjects.
  • the system further includes the logic to make detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects.
  • the system can include third image processors as described herein, including foreground image recognition engines, receiving corresponding sequences of images from the plurality of cameras.
  • the third image processors process images to identify and classify foreground changes represented in the images in the corresponding sequences of images.
  • a system and methods for operating a system are provided for tracking multi- joint subj ects, such as persons, in real space.
  • the system uses a plurality of cameras to produce respective sequences of images of corresponding fields of view in the real space.
  • the field of view of each camera overlaps with the field of view of at least one other camera in the plurality of cameras.
  • the system processes images in the sequences of images to generate arrays of joint data structures corresponding to each image.
  • the arrays of joint data structures corresponding to particular images classify elements of the particular images by the joint type, time of the particular image, and coordinates of the element in the particular image.
  • the system then translates the coordinates of the elements in the arrays of joint data structures corresponding to images in different sequences into candidate joints having coordinates in the real space. Finally, the system identifies constellations of candidate joints, where the constellations include respective sets of candidate joints having coordinates in real space, as multi -joint subjects in the real space.
  • the image recognition engines comprise convolutional neural networks.
  • the processing of images by image recognition engines includes generating confidence arrays for elements of the image.
  • a confidence array for a particular element of an image includes confidence values for a plurality of joint types for the particular element. The confidence arrays are used to select a joint type for the joint data structure of the particular element based on the confidence array.
  • identifying sets of candidate joints comprises applying heuristic functions based on physical relationships among joints of subjects in real space to identify sets of candidate joints as multi -joint subjects.
  • the processing includes storing the sets of joints identified as multi -joint subjects. Identifying sets of candidate joints includes determining whether a candidate joint identified in images taken at a particular time corresponds with a member of one of the sets of candidate joints identified as multi -joint subjects in a preceding image.
  • the sequences of images are synchronized so that images in each of the sequences of images captured by the plurality of cameras represent the real space at a single point in time on the time scale of the movement of subjects through the space.
  • the coordinates in real space of members of a set of candidate joints identified as a multi -joint subject identify locations in the area of the multi -joint subject.
  • the processing includes simultaneous tracking of the locations of a plurality of multi -joint subjects in the area of real space. In some embodiments, the processing includes determining when a multi -joint subject in the plurality of multi -joint subjects leaves the area of real space. In some embodiments, the processing includes determining a direction in which the multi -joint subject is facing at a given point in time. In an embodiment described herein, the system uses a plurality of cameras to produce respective sequences of images of corresponding fields of view in the real space. The field of view of each camera overlaps with the field of view of at least one other camera in the plurality of cameras.
  • the system processes images in the sequences of images received from the plurality of cameras to identify subjects represented in the images and generate classifications of the identified subjects. Finally, the system processes the classifications of identified subjects for sets of images in the sequences of images to detect takes of inventory items by identified subjects and puts of inventory items on shelves by identified subjects.
  • the classification identifies whether the identified subject is holding an inventory item.
  • the classification also identifies whether a hand of the identified subject is near a shelf or whether a hand of the identified subject is near the identified subject.
  • the classification of whether the hand is near the identified subject can include whether a hand of the identified subject is near to a basket associated with an identified subject, and near to the body of the identified subject.
  • the system can make a first set of detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects, and a second set of detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects.
  • Selection logic to process the first and second sets of detections can be used to generate log data structures.
  • the log data structures include lists of inventory items for identified subjects.
  • the sequences of images from cameras in the plurality of cameras are synchronized.
  • the same cameras and the same sequences of images are used by both the foreground and background image processors in one preferred implementation.
  • redundant detections of puts and takes of inventory items are made using the same input data allowing for high confidence, and high accuracy, in the resulting data.
  • the system comprises logic to detect puts and takes of inventory items by identifying gestures of subjects and inventory items associated with the gestures represented in the sequences of images. This can be done using foreground image recognition engines in coordination with subject image recognition engines as described herein.
  • the system comprises logic to detect puts and takes of inventory items by identifying semantically significant changes in inventory items on inventory display structures, such as shelves, over time and associating the semantically significant changes with subjects represented in the sequences of images. This can be done using background image recognition engines in coordination with subject image recognition engines as described herein.
  • gesture analysis and semantic difference analysis can be combined, and executed on the same sequences of synchronized images from an array of cameras.
  • FIG. 1 illustrates an architectural level schematic of a system in which a tracking engine tracks subjects using joint data generated by image recognition engines.
  • FIG. 2 is a side view of an aisle in a shopping store illustrating a camera arrangement.
  • FIG. 3 is a top view of the aisle of Fig. 2 in a shopping store illustrating a camera arrangement.
  • Fig. 4 is a camera and computer hardware arrangement configured for hosting an image recognition engine of Fig. 1.
  • Fig. 5 illustrates a convolutional neural network illustrating identification of joints in an image recognition engine of Fig. 1.
  • Fig. 6 shows an example data structure for storing joint information.
  • Fig. 7 illustrates the tracking engine of Fig. 1 with a global metric calculator.
  • Fig. 8 shows an example data structure for storing a subject including the information of associated joints.
  • Fig. 9 is a flowchart illustrating process steps for tracking subjects by the system of Fig. 1.
  • Fig. 10 is a flowchart showing more detailed process steps for a camera calibration step of Fig. 9.
  • Fig. 11 is a flowchart showing more detailed process steps for a video process step of Fig. 9.
  • Fig. 12A is a flowchart showing a first part of more detailed process steps for the scene process of Fig. 9.
  • Fig. 12B is a flowchart showing a second part of more detailed process steps for the scene process of Fig. 9.
  • FIG. 13 is an illustration of an environment in which an embodiment of the system of Fig. 1 is used.
  • Fig. 14 is an illustration of video and scene processes in an embodiment of the system of Fig. 1.
  • Fig. 15a is a schematic showing a pipeline with multiple convolutional neural networks (CNNs) including joints-CNN, WhatCNN and WhenCNN to generate a shopping cart data structure per subject in the real space.
  • CNNs convolutional neural networks
  • Fig. 15b shows multiple image channels from multiple cameras and coordination logic for the subjects and their respective shopping cart data structures.
  • Fig. 16 is a flowchart illustrating process steps for identifying and updating subjects in the real space.
  • Fig. 17 is a flowchart showing process steps for processing hand joints of subjects to identify inventory items.
  • Fig. 18 is a flowchart showing process steps for time series analysis of inventory items per hand joint to create a shopping cart data structure per subject.
  • FIG. 19 is an illustration of a WhatCNN model in an embodiment of the system of
  • FIG. 20 is an illustration of a WhenCNN model in an embodiment of the system of Fig. 15 a.
  • Fig. 21 presents an example architecture of a WhatCNN model identifying the dimensionality of convolutional layers.
  • Fig. 22 presents a high level block diagram of an embodiment of a WhatCNN model for classification of hand images.
  • FIG. 23 presents details of a first block of the high level block diagram of a
  • Fig. 24 presents operators in a fully connected layer in the example WhatCNN model presented in Fig. 22.
  • Fig. 25 is an example name of an image file stored as part of the training data set for a WhatCN model.
  • Fig. 26 is a high level architecture of a system for tracking changes by subjects in an area of real space in which a selection logic selects between a first detection using background semantic diffing and a redundant detection using foreground region proposals.
  • Fig. 27 presents components of subsystems implementing the system of Fig. 26.
  • Fig. 28A is a flowchart showing a first part of detailed process steps for determining inventory events and generation of the shopping cart data structure.
  • Fig. 28B is a flowchart showing a second part of detailed process steps for determining inventory events and generation of the shopping cart data structure.
  • FIG. 1 An architectural level schematic of a system in accordance with an implementation. Because Fig. 1 is an architectural diagram, certain details are omitted to improve the clarity of the description.
  • FIG. 1 The discussion of Fig. 1 is organized as follows. First, the elements of the system are described, followed by their interconnections. Then, the use of the elements in the system is described in greater detail.
  • Fig. 1 provides a block diagram level illustration of a system 100.
  • the system 100 includes cameras 114, network nodes hosting image recognition engines 1 12a, 1 12b, and 112n, a tracking engine 1 10 deployed in a network node (or nodes) on the network, a calibrator 120, a subject database 140, a training database 150, a heuristics database 160 for joints heuristics, for put and take heuristics, and other heuristics for coordinating and combining the outputs of multiple image recognition engines as described below, a calibration database 170, and a communication network or networks 181.
  • the network nodes can host only one image recognition engine, or several image recognition engines as described herein.
  • the system can also include an inventory database and other supporting data.
  • a network node is an addressable hardware device or virtual device that is attached to a network, and is capable of sending, receiving, or forwarding information over a communications channel to or from other network nodes.
  • Examples of electronic devices which can be deployed as hardware network nodes include all varieties of computers, workstations, laptop computers, handheld computers, and smartphones.
  • Network nodes can be implemented in a cloud-based server system. More than one virtual device configured as a network node can be implemented using a single physical device.
  • network nodes hosting image recognition engines are shown in the system 100. However, any number of network nodes hosting image recognition engines can be connected to the tracking engine 1 10 through the network(s) 181. Also, an image recognition engine, a tracking engine and other processing engines described herein can execute using more than one network node in a distributed architecture.
  • Network(s) 181 couples the network nodes 101a, 101b, and 101 c, respectively, hosting image recognition engines 112a, 112b, and 112n, the network node 102 hosting the tracking engine 1 10, the calibrator 120, the subj ect database 140, the training database 150, the joints heuristics database 160, and the calibration database 170.
  • Cameras 1 14 are connected to the tracking engine 1 10 through network nodes hosting image recognition engines 112a, 112b, and 1 12n.
  • the cameras 114 are installed in a shopping store (such as a supermarket) such that sets of cameras 1 14 (two or more) with overlapping fields of view are positioned over each aisle to capture images of real space in the store.
  • a shopping store such as a supermarket
  • the cameras 114 are installed over aisles with overlapping fields of view.
  • the cameras are configured with the goal that customers moving in the aisles of the shopping store are present in the field of view of two or more cameras at any moment in time.
  • Cameras 114 can be synchronized in time with each other, so that images are captured at the same time, or close in time, and at the same image capture rate.
  • the cameras 114 can send respective continuous streams of images at a predetermined rate to network nodes hosting image recognition engines 1 12a-l 12n. Images captured in all the cameras covering an area of real space at the same time, or close in time, are synchronized in the sense that the synchronized images can be identified in the processing engines as representing different views of subjects having fixed positions in the real space.
  • the cameras send image frames at the rates of 30 frames per second (fps) to respective network nodes hosting image recognition engines 1 12a-l 12n. Each frame has a timestamp, identity of the camera (abbreviated as "camera id”), and a frame identity (abbreviated as "frame id”) along with the image data.
  • Cameras installed over an aisle are connected to respective image recognition engines.
  • the two cameras installed over the aisle 1 16a are connected to the network node 101a hosting an image recognition engine 112a.
  • the two cameras installed over aisle 1 16b are connected to the network node 101b hosting an image recognition engine 112b.
  • Each image recognition engine 112a-l 12n hosted in a network node or nodes 101a- 101 ⁇ separately processes the image frames received from one camera each in the illustrated example.
  • each image recognition engine 112a, 1 12b, and 112n is implemented as a deep learning algorithm such as a convolutional neural network (abbreviated CNN).
  • the CNN is trained using a training database 150.
  • image recognition of subjects in the real space is based on identifying and grouping joints recognizable in the images, where the groups of joints can be attributed to an individual subject.
  • the training database 150 has a large collection of images for each of the different types of joints for subjects.
  • the subjects are the customers moving in the aisles between the shelves.
  • the system 100 is referred to as a "training system”.
  • the CNN After training the CNN using the training database 150, the CNN is switched to production mode to process images of customers in the shopping store in real time.
  • the system 100 is referred to as a runtime system (also referred to as an inference system).
  • the CNN in each image recognition engine produces arrays of joints data structures for images in its respective stream of images.
  • an array of joints data structures is produced for each processed image, so that each image recognition engine 1 12a-l 12n produces an output stream of arrays of joints data structures.
  • These arrays of joints data structures from cameras having overlapping fields of view are further processed to form groups of joints, and to identify such groups of joints as subjects.
  • the cameras 1 14 are calibrated before switching the CNN to production mode.
  • the calibrator 120 calibrates the cameras and stores the calibration data in the calibration database 170.
  • the tracking engine 110 hosted on the network node 102, receives continuous streams of arrays of joints data structures for the subjects from image recognition engines 112a- 1 12n.
  • the tracking engine 110 processes the arrays of joints data structures and translates the coordinates of the elements in the arrays of joints data structures corresponding to images in different sequences into candidate joints having coordinates in the real space.
  • the combination of candidate joints identified throughout the real space can be considered, for the purposes of analogy, to be like a galaxy of candidate joints.
  • movement of the candidate joints is recorded so that the galaxy changes over time.
  • the output of the tracking engine 1 10 is stored in the subject database 140.
  • the tracking engine 110 uses logic to identify groups or sets of candidate joints having coordinates in real space as subjects in the real space.
  • each set of candidate points is like a constellation of candidate joints at each point in time.
  • the constellations of candidate joints can move over time.
  • the logic to identify sets of candidate joints comprises heuristic functions based on physical relationships amongst joints of subjects in real space. These heuristic functions are used to identify sets of candidate joints as subjects.
  • the heuristic functions are stored in heuristics database 160.
  • the output of the tracking engine 1 10 is stored in the subject database 140.
  • the sets of candidate joints comprise individual candidate joints that have relationships according to the heuristic parameters with other individual candidate joints and subsets of candidate joints in a given set that has been identified, or can be identified, as an individual subject.
  • the actual communication path through the network 181 can be point-to-point over public and/or private networks.
  • the communications can occur over a variety of networks 181 , e.g., private networks, VPN, MPLS circuit, or Internet, and can use appropriate application programming interfaces (APIs) and data interchange formats, e.g., Representational State
  • APIs application programming interfaces
  • data interchange formats e.g., Representational State
  • the communication is generally over a network such as a LAN (local area network), WAN (wide area network), telephone network (Public Switched Telephone Network (PSTN), Session Initiation Protocol (SIP), wireless network, point-to-point network, star network, token ring network, hub network, Internet, inclusive of the mobile Internet, via protocols such as EDGE, 3G, 4G LTE, Wi-Fi, and WiMAX.
  • a variety of authorization and authentication techniques such as username/password, Open Authorization (OAuth), Kerberos, SecurelD, digital certificates and more, can be used to secure the communications.
  • the technology disclosed herein can be implemented in the context of any computer-implemented system including a database system, a multi-tenant environment, or a relational database implementation like an OracleTM compatible database implementation, an IBM DB2 Enterprise ServerTM compatible relational database implementation, a MySQLTM or PostgreSQLTM compatible relational database implementation or a Microsoft SQL ServerTM compatible relational database implementation or a NoSQLTM non-relational database implementation such as a VampireTM compatible non-relational database implementation, an Apache CassandraTM compatible non-relational database implementation, a BigTableTM compatible non-relational database implementation or an HBaseTM or DynamoDBTM compatible non-relational database implementation.
  • a relational database implementation like an OracleTM compatible database implementation, an IBM DB2 Enterprise ServerTM compatible relational database implementation, a MySQLTM or PostgreSQLTM compatible relational database implementation or a Microsoft SQL ServerTM compatible relational database implementation or a NoSQLTM non-relational database implementation such as a VampireTM
  • the technology disclosed can be implemented using different programming models like MapReduceTM, bulk synchronous programming, MPI primitives, etc. or different scalable batch and stream management systems like Apache StormTM, Apache SparkTM, Apache KafkaTM, Apache FlinkTM, TruvisoTM, Amazon Elasticsearch ServiceTM, Amazon Web ServicesTM (AWS), IBM Info-SphereTM, BorealisTM, and Yahoo! S4TM.
  • Apache StormTM Apache SparkTM, Apache KafkaTM, Apache FlinkTM, TruvisoTM, Amazon Elasticsearch ServiceTM, Amazon Web ServicesTM (AWS), IBM Info-SphereTM, BorealisTM, and Yahoo! S4TM.
  • the cameras 114 are arranged to track multi -joint entities (or subjects) in a three- dimensional (abbreviated as 3D) real space.
  • the real space can include the area of the shopping store where items for sale are stacked in shelves.
  • a point in the real space can be represented by an (x, y, z) coordinate system.
  • Each point in the area of real space for which the system is deployed is covered by the fields of view of two or more cameras 114.
  • the shelves and other inventory display structures can be arranged in a variety of manners, such as along the walls of the shopping store, or in rows forming aisles or a combination of the two arrangements.
  • Fig. 2 shows an arrangement of shelves, forming an aisle 116a, viewed from one end of the aisle 116a.
  • Two cameras, camera A 206 and camera B 208 are positioned over the aisle 116a at a predetermined distance from a roof 230 and a floor 220 of the shopping store above the inventory display structures such as shelves.
  • the cameras 114 comprise cameras disposed over and having fields of view encompassing respective parts of the inventory display structures and floor area in the real space.
  • the coordinates in real space of members of a set of candidate joints, identified as a subject identify locations in the floor area of the subject.
  • the real space can include all of the floor 220 in the shopping store from which inventory can be accessed.
  • Cameras 114 are placed and oriented such that areas of the floor 220 and shelves can be seen by at least two cameras.
  • the cameras 114 also cover at least part of the shelves 202 and 204 and floor space in front of the shelves 202 and 204.
  • Camera angles are selected to have both steep perspective, straight down, and angled perspectives that give more full body images of the customers.
  • the cameras 114 are configured at an eight (8) foot height or higher throughout the shopping store.
  • Fig. 13 presents an illustration of such an embodiment.
  • the cameras 206 and 208 have overlapping fields of view, covering the space between a shelf A 202 and a shelf B 204 with overlapping fields of view 216 and 218, respectively.
  • a location in the real space is represented as a (x, y, z) point of the real space coordinate system, "x" and "y” represent positions on a two-dimensional (2D) plane which can be the floor 220 of the shopping store.
  • the value "z” is the height of the point above the 2D plane at floor 220 in one configuration.
  • FIG. 3 illustrates the aisle 116a viewed from the top of Fig. 2, further showing an example arrangement of the positions of cameras 206 and 208 over the aisle 116a.
  • the cameras 206 and 208 are positioned closer to opposite ends of the aisle 116a.
  • the camera A 206 is positioned at a predetermined distance from the shelf A 202 and the camera B 208 is positioned at a predetermined distance from the shelf B 204.
  • the cameras are positioned at equal distances from each other.
  • two cameras are positioned close to the opposite ends and a third camera is positioned in the middle of the aisle. It is understood that a number of different camera arrangements are possible.
  • the camera calibrator 120 performs two types of calibrations: internal and external. In internal calibration, the internal parameters of the cameras 114 are calibrated.
  • Examples of internal camera parameters include focal length, principal point, skew, fisheye coefficients, etc.
  • a variety of techniques for internal camera calibration can be used. One such technique is presented by Zhang in "A flexible new technique for camera calibration" published in IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 22, No. 11, November 2000.
  • the external camera parameters are calibrated in order to generate mapping parameters for translating the 2D image data into 3D coordinates in real space.
  • one subject such as a person, is introduced into the real space. The subject moves through the real space on a path that passes through the field of view of each of the cameras 114. At any given point in the real space, the subject is present in the fields of view of at least two cameras forming a 3D scene.
  • the two cameras have a different view of the same 3D scene in their respective two-dimensional (2D) image planes.
  • a feature in the 3D scene such as a left- wrist of the subject is viewed by two cameras at different positions in their respective 2D image planes.
  • a point correspondence is established between every pair of cameras with overlapping fields of view for a given scene. Since each camera has a different view of the same 3D scene, a point correspondence is two pixel locations (one location from each camera with overlapping field of view) that represent the projection of the same point in the 3D scene. Many point correspondences are identified for each 3D scene using the results of the image recognition engines 112a-l 12n for the purposes of the external calibration.
  • the image recognition engines identify the position of a joint as (x, y) coordinates, such as row and column numbers, of pixels in the 2D image planes of respective cameras 114.
  • a joint is one of 19 different types of joints of the subject. As the subject moves through the fields of view of different cameras, the tracking engine 110 receives (x, y) coordinates of each of the 19 different types of joints of the subject used for the calibration from cameras 1 14 per image.
  • images are streamed off of all cameras at a rate of 30 FPS (frames per second) or more and a resolution of 720 pixels in full RGB (red, green, and blue) color. These images are in the form of one-dimensional arrays (also referred to as flat arrays).
  • FPS frames per second
  • RGB red, green, and blue
  • These images are in the form of one-dimensional arrays (also referred to as flat arrays).
  • the large number of images collected above for a subject can be used to determine corresponding points between cameras with overlapping fields of view. Consider two cameras A and B with overlapping field of view. The plane passing through camera centers of cameras A and B and the joint location (also referred to as feature point) in the 3D scene is called the "epipolar plane".
  • This non-linear optimization function is used by the tracking engine 1 10 to identify the same joints in outputs (arrays of joints data structures) of different image recognition engines 1 12a-112n, processing images of cameras 114 with overlapping fields of view.
  • the results of the internal and external camera calibration are stored in the calibration database 170.
  • a variety of techniques for determining the relative positions of the points in images of cameras 1 14 in the real space can be used.
  • Longuet-Higgins published, "A computer algorithm for reconstructing a scene from two projections" in Nature, Volume 293, 10 September 1981. This paper presents computing a three-dimensional structure of a scene from a correlated pair of perspective projections when spatial relationship between the two projections is unknown.
  • the Longuet-Higgins paper presents a technique to determine the position of each camera in the real space with respect to other cameras. Additionally, their technique allows triangulation of a subject in the real space, identifying the value of the z-coordinate (height from the floor) using images from cameras 1 14 with overlapping fields of view.
  • An arbitrary point in the real space for example, the end of a shelf in one corner of the real space, is designated as a (0, 0, 0) point on the (x, y, z) coordinate system of the real space.
  • the parameters of the external calibration are stored in two data structures.
  • the first data structure stores intrinsic parameters.
  • the intrinsic parameters represent a projective transformation from the 3D coordinates into 2D image coordinates.
  • the first data structure contains intrinsic parameters per camera as shown below.
  • the data values are all numeric floating point numbers.
  • This data structure stores a 3x3 intrinsic matrix, represented as "K" and distortion coefficients.
  • the distortion coefficients include six radial distortion coefficients and two tangential distortion coefficients. Radial distortion occurs when light rays bend more near the edges of a lens than they do at its optical center. Tangential distortion occurs when the lens and the image plane are not parallel.
  • the following data structure shows values for the first camera only. Similar data is stored for all the cameras 114.
  • the second data structure stores per pair of cameras: a 3x3 fundamental matrix
  • F a 3x3 essential matrix
  • E a 3x4 projection matrix
  • R a 3x3 rotation matrix
  • t a 3x1 translation vector
  • This data is used to convert points in one camera's reference frame to another camera's reference frame.
  • eight homography coefficients are also stored to map the plane of the floor 220 from one camera to another.
  • a fundamental matrix is a relationship between two images of the same scene that constrains where the projection of points from the scene can occur in both images.
  • Essential matrix is also a relationship between two images of the same scene with the condition that the cameras are calibrated.
  • the projection matrix gives a vector space projection from 3D real space to a subspace.
  • the rotation matrix is used to perform a rotation in Euclidean space.
  • Translation vector "t” represents a geometric transformation that moves every point of a figure or a space by the same distance in a given direction.
  • the homography floor coefficients are used to combine images of features of subjects on the floor 220 viewed by cameras with overlapping fields of views.
  • the second data structure is shown below. Similar data is stored for all pairs of cameras. As indicated previously, the x's represents numeric floating point numbers.
  • homography floor coefficients [x, x, x, x, x, x, x]
  • Fig. 4 presents an architecture 400 of a network hosting image recognition engines.
  • the system includes a plurality of network nodes l Ol a-l Oln in the illustrated embodiment.
  • the network nodes are also referred to as processing platforms.
  • Processing platforms l Ol a-lOln and cameras 412, 414, 416, ... 418 are connected to network(s) 481.
  • Fig. 4 shows a plurality of cameras 412, 414, 416, ... 418 connected to the network(s).
  • the cameras 412 to 418 are connected to the network(s) 481 using Ethernet-based connectors 422, 424, 426, and 428, respectively.
  • the Ethernet-based connectors have a data transfer speed of 1 gigabit per second, also referred to as Gigabit Ethernet.
  • cameras 114 are connected to the network using other types of network connections which can have a faster or slower data transfer rate than Gigabit Ethernet.
  • a set of cameras can be connected directly to each processing platform, and the processing platforms can be coupled to a network.
  • Storage subsystem 430 stores the basic programming and data constructs that provide the functionality of certain embodiments of the present invention.
  • the various modules implementing the functionality of a plurality of image recognition engines may be stored in storage subsystem 430.
  • the storage subsystem 430 is an example of a computer readable memory comprising a non-transitory data storage medium, having computer instructions stored in the memory executable by a computer to perform the all or any combination of the data processing and image processing functions described herein, including logic to identify changes in real space, to track subjects and to detect puts and takes of inventory items in an area of real space by processes as described herein.
  • the computer instructions can be stored in other types of memory, including portable memory, that comprise a non-transitory data storage medium or media, readable by a computer.
  • a host memory subsystem 432 typically includes a number of memories including a main random access memory (RAM) 434 for storage of instructions and data during program execution and a read-only memory (ROM) 436 in which fixed instructions are stored.
  • RAM main random access memory
  • ROM read-only memory
  • the RAM 434 is used as a buffer for storing video streams from the cameras 114 connected to the platform 101 a.
  • a file storage subsystem 440 provides persistent storage for program and data files.
  • the storage subsystem 440 includes four 120 Gigabyte (GB) solid state disks (SSD) in a RAID 0 (redundant array of independent disks) arrangement identified by a numeral 442.
  • the RAID 0 442 is used to store training data.
  • the training data which is not in RAM 434 is read from RAID 0 442.
  • the hard disk drive (HDD) 446 is a 10 terabyte storage. It is slower in access speed than the RAID 0 442 storage.
  • the solid state disk (SSD) 444 contains the operating system and related files for the image recognition engine 1 12a.
  • each camera has a dedicated graphics processing unit GPU 1 462, GPU 2 464, and GPU 3 466, to process images sent by the camera. It is understood that fewer than or more than three cameras can be connected per processing platform. Accordingly, fewer or more GPUs are configured in the network node so that each camera has a dedicated GPU for processing the image frames received from the camera.
  • the processor subsystem 450, the storage subsystem 430 and the GPUs 462 , 464, and 466 communicate using the bus subsystem 454.
  • a number of peripheral devices such as a network interface subsystem, user interface output devices, and user interface input devices are also connected to the bus subsystem 454 forming part of the processing platform 101a. These subsystems and devices are
  • bus subsystem 454 is shown schematically as a single bus, alternative embodiments of the bus subsystem may use multiple busses.
  • the cameras 412 can be implemented using Chameleon3 1.3
  • MP Color USB3 Vision (Sony ICX445), having a resolution of 1288 x 964, a frame rate of 30 FPS, and at 1.3 Megapixels per image, with Varifocal Lens having a working distance (mm) of 300 - ⁇ , a field of view field of view with a 1/3" sensor of 98.2° - 23.8°.
  • the image recognition engines in the processing platforms receive a continuous stream of images at a predetermined rate.
  • the image recognition engines comprise convolutional neural networks (abbreviated CNN).
  • FIG. 5 illustrates processing of image frames by a CNN referred to by a numeral
  • the input image 510 is a matrix consisting of image pixels arranged in rows and columns.
  • the input image 510 has a width of 1280 pixels, height of 720 pixels and 3 channels red, blue, and green also referred to as RGB.
  • the channels can be imagined as three 1280 x 720 two-dimensional images stacked over one another. Therefore, the input image has dimensions of 1280 x 720 x 3 as shown in Fig. 5.
  • a 2 x 2 filter 520 is convolved with the input image 510.
  • no padding is applied when the filter is convolved with the input.
  • a nonlinearity function is applied to the convolved image.
  • rectified linear unit (ReLU) activations are used.
  • Other examples of nonlinear functions include sigmoid, hyperbolic tangent (tanh) and variations of ReLU such as leaky ReLU.
  • a search is performed to find hyper- parameter values.
  • the hyper-parameters are C 1; C 2 , , CN where CN means the number of channels for convolution layer "N". Typical values of N and C are shown in Fig. 5.
  • N there are twenty five (25) layers in the CNN as represented by N equals 25.
  • the values of C are the number of channels in each convolution layer for layers 1 to 25.
  • additional features are added to the CNN 500 such as residual connections, squeeze-excitation modules, and multiple resolutions.
  • the size of the image is reduced as the image is processed through convolution layers. That is helpful in feature identification as the goal is to predict a class for the input image.
  • the size of the input image i.e. image width and height dimensions
  • the size of the input image is not reduced, as the goal is to not only to identify a joint (also referred to as a feature) in the image frame, but also to identify its location in the image so it can be mapped to coordinates in the real space. Therefore, as shown Fig. 5, the width and height dimensions of the image remain unchanged as the processing proceeds through convolution layers of the CNN, in this example.
  • the CNN 500 identifies one of the 19 possible joints of the subjects at each element of the image.
  • the possible joints can be grouped in two categories: foot joints and non-foot joints.
  • the 19 type of joint classification is for all non-joint features of the subject (i.e. elements of the image not classified as a joint).
  • a "joint" for the purposes of this description is a trackable feature of a subject in the real space.
  • a joint may correspond to physiological joints on the subjects, or other features such as the eye, or nose.
  • the first set of analyses on the stream of input images identifies trackable features of subjects in real space. In one embodiment, this is referred to as “joints analysis”. In such an embodiment, the CNN used for joints analysis is referred to as “joints CNN”. In one
  • the joints analysis is performed thirty times per second over thirty frames per second received from the corresponding camera.
  • the analysis is synchronized in time i.e., at l/30 th of a second, images from all cameras 114 are analyzed in the corresponding joints CNNs to identify joints of all subjects in the real space.
  • the results of this analysis of the images from a single moment in time from plural cameras is stored as a "snapshot".
  • a snapshot can be in the form of a dictionary containing arrays of joints data structures from images of all cameras 114 at a moment in time, representing a constellation of candidate joints within the area of real space covered by the system.
  • the snapshot is stored in the subject database 140.
  • a softmax function is applied to every element of the image in the final layer of convolution layers 530.
  • the softmax function transforms a i ⁇ -dimensional vector of arbitrary real values to a i ⁇ -dimensional vector of real values in the range [0, 1] that add up to 1.
  • an element of an image is a single pixel.
  • the softmax function converts the 19-dimensional array (also referred to a 19-dimensional vector) of arbitrary real values for each pixel to a 19-dimensional confidence array of real values in the range [0, 1] that add up to 1.
  • the 19 dimensions of a pixel in the image frame correspond to the 19 channels in the final layer of the CNN which further correspond to 19 types of joints of the subj ects.
  • a large number of picture elements can be classified as one of each of the 19 types of joints in one image depending on the number of subjects in the field of view of the source camera for that image.
  • the image recognition engines 112a- 1 12n process images to generate confidence arrays for elements of the image.
  • a confidence array for a particular element of an image includes confidence values for a plurality of joint types for the particular element.
  • Each one of the image recognition engines 1 12a-112n, respectively, generates an output matrix 540 of confidence arrays per image.
  • each image recognition engine generates arrays of joints data structures corresponding to each output matrix 540 of confidence arrays per image.
  • the arrays of joints data structures corresponding to particular images classify elements of the particular images by joint type, time of the particular image, and coordinates of the element in the particular image.
  • a joint type for the joints data structure of the particular elements in each image is selected based on the values of the confidence array.
  • the heat map can be resolved to show image elements having the highest values (peak) for each joint type.
  • peak the highest values for each joint type.
  • surrounding picture elements outside a range from the given picture element will have lower values for that joint type, so that a location for a particular joint having that joint type can be identified in the image space coordinates.
  • the confidence array for that image element will have the highest confidence value for that joint and lower confidence values for the remaining 18 types of joints.
  • batches of images from each camera 1 14 are processed by respective image recognition engines. For example, six contiguously timestamped images are processed sequentially in a batch to take advantage of cache coherence.
  • the parameters for one layer of the CNN 500 are loaded in memory and applied to the batch of six image frames. Then the parameters for the next layer are loaded in memory and applied to the batch of six images. This is repeated for all convolution layers 530 in the CNN 500.
  • the cache coherence reduces processing time and improves performance of the image recognition engines.
  • 3D convolution a further improvement in performance of the CNN 500 is achieved by sharing information across image frames in the batch. This helps in more precise identification of joints and reduces false positives.
  • features in the image frames for which pixel values do not change across the multiple image frames in a given batch are likely static objects such as a shelf.
  • the change of values for the same pixel across image frames in a given batch indicates that this pixel is likely a joint. Therefore, the CNN 500 can focus more on processing that pixel to accurately identify the joint identified by that pixel.
  • the output of the CNN 500 is a matrix of confidence arrays for each image per camera.
  • the matrix of confidence arrays is transformed into an array of joints data structures.
  • a joints data structure 600 as shown in Fig. 6 is used to store the information of each joint.
  • the joints data structure 600 identifies x and y positions of the element in the particular image in the 2D image space of the camera from which the image is received.
  • a j oint number identifies the type of joint identified. For example, in one embodiment, the values range from 1 to 19.
  • a value of 1 indicates that the joint is a left-ankle
  • a value of 2 indicates the joint is a right-ankle and so on.
  • the type of joint is selected using the confidence array for that element in the output matrix 540. For example, in one embodiment, if the value corresponding to the left-ankle j oint is highest in the confidence array for that image element, then the value of the joint number is "1 ".
  • a confidence number indicates the degree of confidence of the CNN 500 in predicting that joint. If the value of confidence number is high, it means the CNN is confident in its prediction.
  • An integer-Id is assigned to the joints data structure to uniquely identify it.
  • the output matrix 540 of confidence arrays per image is converted into an array of joints data structures for each image.
  • the image recognition engines 1 12a-l 12n receive the sequences of images from cameras 1 14 and process images to generate corresponding arrays of joints data structures as described above.
  • An array of joints data structures for a particular image classifies elements of the particular image by joint type, time of the particular image, and the coordinates of the elements in the particular image.
  • the image recognition engines 112a-l 12n are convolutional neural networks CNN 500, the joint type is one of the 19 types of joints of the subjects, the time of the particular image is the timestamp of the image generated by the source camera 1 14 for the particular image, and the coordinates (x, y) identify the position of the element on a 2D image plane.
  • the joints analysis includes performing a combination of k- nearest neighbors, mixture of Gaussians, various image morphology transformations, and joints CNN on each input image.
  • the result comprises arrays of joints data structures which can be stored in the form of a bit mask in a ring buffer that maps image numbers to bit masks at each moment in time.
  • the tracking engine 110 is configured to receive arrays of joints data structures generated by the image recognition engines 1 12a-l 12n corresponding to images in sequences of images from cameras having overlapping fields of view.
  • the arrays of joints data structures per image are sent by image recognition engines 112a-l 12n to the tracking engine 110 via the network(s) 181 as shown in Fig. 7.
  • the tracking engine 110 translates the coordinates of the elements in the arrays of joints data structures corresponding to images in different sequences into candidate joints having coordinates in the real space.
  • the tracking engine 110 comprises logic to identify sets of candidate joints having coordinates in real space (constellations of joints) as subjects in the real space.
  • the tracking engine 110 accumulates arrays of joints data structures from the image recognition engines for all the cameras at a given moment in time and stores this information as a dictionary in the subject database 140, to be used for identifying a constellation of candidate joints.
  • the dictionary can be arranged in the form of key- value pairs, where keys are camera ids and values are arrays of joints data structures from the camera.
  • this dictionary is used in heuristics-based analysis to determine candidate joints and for assignment of joints to subjects.
  • a high-level input, processing and output of the tracking engine 1 10 is illustrated in table 1.
  • Table 1 Inputs, processing and outputs from tracking engine 110 in an example embodiment.
  • the tracking engine 1 10 receives arrays of joints data structures along two dimensions: time and space. Along the time dimension, the tracking engine receives sequentially timestamped arrays of joints data structures processed by image recognition engines 1 12a-l 12n per camera.
  • the joints data structures include multiple instances of the same joint of the same subject over a period of time in images from cameras having overlapping fields of view.
  • the (x, y) coordinates of the element in the particular image will usually be different in sequentially timestamped arrays of joints data structures because of the movement of the subject to which the particular joint belongs. For example, twenty picture elements classified as left- wrist joints can appear in many sequentially timestamped images from a particular camera, each left- wrist joint having a position in real space that can be changing or unchanging from image to image. As a result, twenty left- wrist joints data structures 600 in many sequentially timestamped arrays of joints data structures can represent the same twenty joints in real space over time.
  • the tracking engine 1 10 receives joints data structures for a particular joint from multiple cameras having overlapping fields of view, at any given moment in time. This is the space dimension, the second of the two dimensions: time and space, along which the tracking engine 110 receives data in arrays of joints data structures.
  • the tracking engine 110 uses an initial set of heuristics stored in the heuristics database 160 to identify candidate joints data structures from the arrays of joints data structures. The goal is to minimize a global metric over a period of time.
  • a global metric calculator 702 calculates the global metric.
  • the global metric is a summation of multiple values described below. Intuitively, the value of the global metric is minimum when the joints in arrays of joints data structures received by the tracking engine 110 along the time and space dimensions are correctly assigned to respective subjects. For example, consider the embodiment of the shopping store with customers moving in the aisles. If the left-wrist of a customer A is incorrectly assigned to a customer B, then the value of the global metric will increase. Therefore, minimizing the global metric for each joint for each customer is an optimization problem. One option to solve this problem is to try all possible connections of joints. However, this can become intractable as the number of customers increases.
  • a second approach to solve this problem is to use heuristics to reduce possible combinations of joints identified as members of a set of candidate joints for a single subject. For example, a left- wrist joint cannot belong to a subject far apart in space from other joints of the subject because of known physiological characteristics of the relative positions of joints.
  • a left- wrist joint having a small change in position from image to image is less likely to belong to a subject having the same joint at the same position from an image far apart in time, because the subjects are not expected to move at a very high speed.
  • These initial heuristics are used to build boundaries in time and space for constellations of candidate joints that can be classified as a particular subject.
  • the joints in the joints data structures within a particular time and space boundary are considered as "candidate joints" for assignment to sets of candidate joints as subjects present in the real space.
  • These candidate joints include joints identified in arrays of joints data structures from multiple images from a same camera over a period of time (time dimension) and across different cameras with overlapping fields of view (space dimension).
  • the joints can be divided for the purposes of a procedure for grouping the joints into constellations, into foot and non-foot joints as shown above in the list of joints.
  • the left and right-ankle joint types in the current example are considered foot joints for the purpose of this procedure.
  • the tracking engine 110 can start identification of sets of candidate joints of particular subjects using foot joints.
  • the feet of the customers are on the floor 220 as shown in Fig. 2.
  • the distance of the cameras 114 to the floor 220 is known. Therefore, when combining the joints data structures of foot joints from arrays of data joints data structures corresponding to images of cameras with overlapping fields of view, the tracking engine 110 can assume a known depth (distance along z axis).
  • the value depth for foot joints is zero i.e. (x, y, 0) in the (x, y, z) coordinate system of the real space.
  • the image tracking engine 110 applies homographic mapping to combine joints data structures of foot joints from cameras with overlapping fields of view to identify the candidate foot joint.
  • the location of the j oint in (x, y) coordinates in image space is converted to the location in the (x, y, z) coordinates in the real space, resulting in a candidate foot joint. This process is performed separately to identify candidate left and right foot joints using respective joints data structures.
  • the tracking engine 110 can combine a candidate left foot joint and a candidate right foot j oint (assigns them to a set of candidate joints) to create a subject.
  • Other joints from the galaxy of candidate joints can be linked to the subject to build a constellation of some or all of the joint types for the created subject.
  • the tracking engine 1 10 creates a new subject having the left and the right candidate foot joints belonging to its set of joints. The subject is saved in the subject database 140. If there are multiple candidate left and right foot joints, then the global metric calculator 702 attempts to combine each candidate left foot joint to each candidate right foot joint to create subjects such that the value of the global metric is minimized.
  • the tracking engine 110 uses the non-linear transformation (also referred to as a fundamental matrix) from any given camera A to its neighboring camera B with overlapping fields of view.
  • the non-linear transformations are calculated using a single multi -joint subject and stored in the calibration database 170 as described above.
  • the candidate non-foot joints are identified as follows.
  • the non-foot joints in arrays of joints data structures corresponding to elements in image frames from camera A are mapped to epipolar lines in synchronized image frames from camera B.
  • a joint (also referred to as a feature in machine vision literature) identified by a joints data structure in an array of joints data structures of a particular image of camera A will appear on a corresponding epipolar line if it appears in the image of camera B.
  • a joint in the joints data structure from camera A is a left- wrist joint
  • a left- wrist joint on the epipolar line in the image of camera B represents the same left- wrist joint from the perspective of camera B.
  • Machine vision techniques such as the technique by Longuet-Higgins published in the paper, titled, "A computer algorithm for reconstructing a scene from two projections" in Nature, Volume 293, 10 September 1981 , are applied to conjugate pairs of corresponding points to determine height of joints from the floor 220 in the real space.
  • Application of the above method requires predetermined mapping between cameras with overlapping fields of view. That data is stored in the calibration database 170 as non-linear functions determined during the calibration of the cameras 1 14 described above.
  • the tracking engine 110 receives the arrays of joints data structures
  • the identified candidate non-foot joints are grouped into sets of subjects having coordinates in real space using the global metric calculator 702.
  • the global metric calculator 702 calculates the global metric value and attempts to minimize the value by checking different combinations of non-foot joints.
  • the global metric is a sum of heuristics organized in four categories.
  • the logic to identify sets of candidate joints comprises heuristic functions based on physical relationships among joints of subjects in real space to identify sets of candidate joints as subjects. Examples of physical relationships among joints are considered in the heuristics as described below.
  • the first category of heuristics includes metrics to ascertain similarity between two proposed subject-joint locations in the same camera view at the same or different moments in time. In one embodiment, these metrics are floating point values, where higher values mean two lists of joints are likely to belong to the same subject.
  • the metrics determine the distance between a customer's same joints in one camera from one image to the next image along the time dimension. Given a customer A in the field of view of the camera 412, the first set of metrics determines the distance between each of person A's joints from one image from the camera 412 to the next image from the camera 412.
  • the metrics are applied to joints data structures 600 in arrays of joints data structures per image from cameras 1 14.
  • two example metrics in the first category of heuristics are listed below: 1.
  • the second category of heuristics includes metrics to ascertain similarity between two proposed subject-joint locations from the fields of view of multiple cameras at the same moment in time.
  • these metrics are floating point values, where higher values mean two lists of joints are likely to belong to the same subject.
  • the second set of metrics determines the distance between a customer's same joints in image frames from two or more cameras (with overlapping fields of view) at the same moment in time.
  • the third category of heuristics include metrics to ascertain similarity between all joints of a proposed subject-j oint location in the same camera view at the same moment in time.
  • this category of metrics determines distance between joints of a customer in one frame from one camera.
  • the fourth category of heuristics includes metrics to ascertain dissimilarity between proposed subject-joint locations.
  • these metrics are floating point values. Higher values mean two lists of joints are more likely to not be the same subject.
  • two example metrics in this category include:
  • Thresholds to determine when there are too many potential candidate subjects that a joint can belong to with too good of a metric similarity score.
  • Thresholds to determine when collections of joints over time have high enough metric similarity to be considered a new subject, previously not present in the real space.
  • Thresholds to determine when a subject is no longer in the real space.
  • Thresholds to determine when the tracking engine 110 has made a mistake and has confused two subjects.
  • the tracking engine 110 includes logic to store the sets of joints identified as subjects.
  • the logic to identify sets of candidate joints includes logic to determine whether a candidate joint identified in images taken at a particular time corresponds with a member of one of the sets of candidate joints identified as subjects in preceding images.
  • the tracking engine 110 compares the current joint-locations of a subject with previously recorded joint-locations of the same subject at regular intervals. This comparison allows the tracking engine 110 to update the joint locations of subjects in the real space. Additionally, using this, the tracking engine 110 identifies false positives (i.e., falsely identified subjects) and removes subjects no longer present in the real space.
  • the tracking engine 110 creates a customer (subject) at an earlier moment in time, however, after some time, the tracking engine 110 does not have current joint-locations for that particular customer. It means that the customer was incorrectly created.
  • the tracking engine 110 deletes incorrectly generated subjects from the subject database 140.
  • the tracking engine 110 also removes positively identified subjects from the real space using the above described process.
  • the tracking engine 110 updates this customer's record in the subject database 140 to indicate that "customer has left the store".
  • the tracking engine 110 attempts to identify subjects by applying the foot and non-foot heuristics simultaneously. This results in "islands" of connected joints of the subjects. As the tracking engine 110 processes further arrays of joints data structures along the time and space dimensions, the size of the islands increases. Eventually, the islands of joints merge to other islands of joints forming subjects which are then stored in the subject database 140. In one embodiment, the tracking engine 110 maintains a record of unassigned joints for a predetermined period of time. During this time, the tracking engine attempts to assign the unassigned joint to existing subjects or create new multi -joint entities from these unassigned joints. The tracking engine 110 discards the unassigned joints after a predetermined period of time. It is understood that, in other embodiments, different heuristics than the ones listed above are used to identify and track subjects.
  • a user interface output device connected to the node 102 hosting the tracking engine 110 displays position of each subject in the real spaces.
  • the display of the output device is refreshed with new locations of the subjects at regular intervals.
  • Fig. 8 shows the subject data structure 800 to store the subject.
  • the data structure 800 stores the subject related data as a key- value dictionary.
  • the key is a frame number and value is another key-value dictionary where key is the camera id and value is a list of 18 joints (of the subject) with their locations in the real space.
  • the subject data is stored in the subject database 140. Every new subject is also assigned a unique identifier that is used to access the subject's data in the subject database 140.
  • the system identifies joints of a subject and creates a skeleton of the subject.
  • the skeleton is projected into the real space indicating the position and orientation of the subject in the real space. This is also referred to as "pose estimation" in the field of machine vision.
  • the system displays orientations and positions of subjects in the real space on a graphical user interface (GUI).
  • GUI graphical user interface
  • the image analysis is anonymous, i.e., a unique identifier assigned to a subject created through joints analysis does not identify personal identification details (such as names, email addresses, mailing addresses, credit card numbers, bank account numbers, driver's license number, etc.) of any specific subject in the real space.
  • a number of flowcharts illustrating logic are described herein.
  • the logic can be implemented using processors configured as described above programmed using computer programs stored in memory accessible and executable by the processors, and in other configurations, by dedicated logic hardware, including field programmable integrated circuits, and by combinations of dedicated logic hardware and computer programs.
  • all flowcharts herein it will be appreciated that many of the steps can be combined, performed in parallel, or performed in a different sequence, without affecting the functions achieved. In some cases, as the reader will appreciate, a rearrangement of steps will achieve the same results only if certain other changes are made as well. In other cases, as the reader will appreciate, a rearrangement of steps will achieve the same results only if certain conditions are satisfied.
  • the flow charts herein show only steps that are pertinent to an understanding of the embodiments, and it will be understood that numerous additional steps for accomplishing other functions can be performed before, after and between those shown.
  • Fig. 9 is a flowchart illustrating process steps for tracking subjects.
  • the process starts at step 902.
  • the cameras 114 having field of view in an area of the real space are calibrated in process step 904.
  • Video processes are performed at step 906 by image recognition engines
  • the video process is performed per camera to process batches of image frames received from respective cameras.
  • the output of all video processes from respective image recognition engines 112a-l 12n are given as input to a scene process performed by the tracking engine 110 at step 908.
  • the scene process identifies new subjects and updates the joint locations of existing subjects.
  • it is checked if there are more image frames to be processed. If there are more image frames, the process continues at step 906, otherwise the process ends at step 914.
  • step 904 More detailed process steps of the process step 904 "calibrate cameras in real space" are presented in a flowchart in Fig. 10.
  • the calibration process starts at step 1002 by identifying a (0, 0, 0) point for (x, y, z) coordinates of the real space.
  • step 1004 a first camera with the location (0, 0, 0) in its field of view is calibrated. More details of camera calibration are presented earlier in this application.
  • step 1006 a next camera with overlapping field of view with the first camera is calibrated.
  • step 1008 it is checked whether there are more cameras to calibrate. The process is repeated at step 1006 until all cameras 114 are calibrated.
  • a subject is introduced in the real space to identify conjugate pairs of corresponding points between cameras with overlapping fields of view. Some details of this process are described above. The process is repeated for every pair of overlapping cameras at step 1012. The process ends if there are no more cameras (step 1014).
  • FIG. 1 1 shows more detailed steps of the "video process” step 906.
  • k-contiguously timestamped images per camera are selected as a batch for further processing.
  • the value of k 6 which is calculated based on available memory for the video process in the network nodes lOla-lOln, respectively hosting image recognition engines 112a-l 12n.
  • the size of images is set to appropriate dimensions. In one embodiment, the images have a width of 1280 pixels, height of 720 pixels and three channels RGB (representing red, green and blue colors).
  • a plurality of trained convolutional neural networks (CNN) process the images and generate arrays of joints data structures per image.
  • the output of the CNNs are arrays of joints data structures per image (step 1 108). This output is sent to a scene process at step 11 10.
  • Fig. 12A is a flowchart showing a first part of more detailed steps for "scene process" step 908 in Fig. 9.
  • the scene process combines outputs from multiple video processes at step 1202.
  • heuristics are applied on candidate foot joints identified in step 1206 to identify sets of candidate foot joints as subjects. It is checked at step 1210 whether the set of candidate foot joints belongs to an existing subject. If not, a new subject is created at step 1212. Otherwise, the existing subj ect is updated at step 1214.
  • a flowchart Fig. 12B illustrates a second part of more detailed steps for the
  • step 1240 the data structures of non-foot joints are combined from multiple arrays of joints data structures corresponding to images in the sequence of images from cameras with overlapping fields of view. This is performed by mapping corresponding points from a first image from a first camera to a second image from a second camera with overlapping fields of view. Some details of this process are described above. Heuristics are applied at step 1242 to candidate non-foot joints. At step 1246 it is determined whether a candidate non-foot joint belongs to an existing subject. If so, the existing subject is updated at step 1248. Otherwise, the candidate non-foot joint is processed again at step 1250 after a predetermined time to match it with an existing subject. At step 1252 it is checked whether the non-foot joint belongs to an existing subject. If true, the subject is updated at step 1256. Otherwise, the joint is discarded at step 1254.
  • the processes to identify new subjects, track subjects and eliminate subjects are implemented as part of an "entity cohesion algorithm" performed by the runtime system (also referred to as the inference system).
  • An entity is a constellation of joints referred to as subject above.
  • the entity cohesion algorithm identifies entities in the real space and updates locations of the joints in real space to track movement of the entity.
  • Fig. 14 presents an illustration of video processes 141 1 and scene process 1415.
  • four video processes are shown, each processing images from one or more cameras 1 14.
  • the video processes process images as described above and identify joints per frame.
  • each video process identifies the 2D coordinates, a confidence number, a joint number and a unique ID per joint per frame.
  • the outputs 1452 of all video processes are given as input 1453 to the scene process 1415.
  • the scene process creates a joint key-value dictionary per moment in time in which the key is the camera identifier and the value is the arrays of joints.
  • the j oints are re-projected into perspectives of cameras with overlapping fields of view.
  • the re-projected joints are stored as a key-value dictionary, and can be used to produce foreground subject masks for each image in each camera as discussed below.
  • the key in this dictionary is a combination of joint id and camera id.
  • the values in the dictionary are 2D coordinates of the joint re-proj ected into the target camera's perspective.
  • the scene process 1415 produces an output 1457 comprising a list of all subjects in the real space at a moment in time.
  • the list includes a key-value dictionary per subject.
  • the key is a unique identifier of a subject and the value is another key-value dictionary with the key as the frame number and the value as the camera-subject joint key-value dictionary.
  • the camera- subject joint key-value dictionary is a per subject dictionary in which the key is the camera identifier and the value is a list of joints.
  • FIG. 15A An architectural level schematic of a system in accordance with an implementation. Because Fig. 15 A is an architectural diagram, certain details are omitted to improve the clarity of the description.
  • Fig. 15 A is a high-level architecture of pipelines of convolutional neural networks
  • multi-CNN pipelines processing image frames received from cameras 1 14 to generate shopping cart data structures for each subject in the real space.
  • the system described here includes per camera image recognition engines as described above for identifying and tracking multi -joint subj ects.
  • Alternative image recognition engines can be used, including examples in which only one "joint" is recognized and tracked per individual, or other features or other types of images data over space and time are utilized to recognize and track subjects in the real space being processed.
  • the multi-CNN pipelines run in parallel per camera, moving images from respective cameras to image recognition engines 112a-112n via circular buffers 1502 per camera.
  • the system is comprised of three subsystems: first image processors subsystem 2602, second image processors subsystem 2604 and third image processors subsystem 2606.
  • the first image processors subsystem 2602 includes image recognition engines 112a-l 12n implemented as convolutional neural networks (CNNs) and referred to as joint CNNs 112a-112n.
  • CNNs convolutional neural networks
  • cameras 114 can be synchronized in time with each other, so that images are captured at the same time, or close in time, and at the same image capture rate. Images captured in all the cameras covering an area of real space at the same time, or close in time, are synchronized in the sense that the synchronized images can be identified in the processing engines as representing different views at a moment in time of subjects having fixed positions in the real space.
  • the cameras 1 14 are installed in a shopping store (such as a supermarket) such that sets of cameras (two or more) with overlapping fields of view are positioned over each aisle to capture images of real space in the store.
  • a shopping store such as a supermarket
  • sets of cameras two or more with overlapping fields of view are positioned over each aisle to capture images of real space in the store.
  • Each camera produces a sequence of images of real space corresponding to its respective field of view.
  • the image frames corresponding to sequences of images from each camera are sent at the rate of 30 frames per second (fps) to respective image recognition engines 112a-l 12n.
  • Each image frame has a timestamp, identity of the camera (abbreviated as "camera id”), and a frame identity (abbreviated as "frame id") along with the image data.
  • the image frames are stored in a circular buffer 1502 (also referred to as a ring buffer) per camera 114.
  • Circular buffers 1502 store a set of consecutively timestamped image frames from respective cameras 114.
  • a joints CNN processes sequences of image frames per camera and identifies 18 different types of joints of each subject present in its respective field of view.
  • the outputs of joints CNNs 112a-l 12n corresponding to cameras with overlapping fields of view are combined to map the location of joints from 2D image coordinates of each camera to 3D coordinates of real space.
  • the joints data structures 800 per subject (j) where j equals 1 to x, identify locations of joints of a subject (j) in the real space.
  • the details of subject data structure 800 are presented in Fig. 8.
  • the joints data structure 800 is a two level key -value dictionary of joints of each subject.
  • a first key is the frame number and the value is a second key-value dictionary with the key as the camera id and the value as the list of joints assigned to a subject.
  • the data sets comprising subjects identified by joints data structures 800 and corresponding image frames from sequences of image frames per camera are given as input to a bounding box generator 1504 in the third image processors subsystem 2606.
  • the third image processors subsystem further comprise foreground image recognition engines.
  • the foreground image recognition engines recognize semantically significant objects in the foreground (i.e. shoppers, their hands and inventory items) as they relate to puts and takes of inventory items for example, over time in the images from each camera.
  • the foreground image recognition engines are implemented as WhatCNN 1506 and WhenCNN 1508.
  • the bounding box generator 1504 implements the logic to process the data sets to specify bounding boxes which include images of hands of identified subjects in images in the sequences of images.
  • the bounding box generator 1504 identifies locations of hand joints in each source image frame per camera using locations of hand joints in the multi -joints data structures 800 corresponding to the respective source image frame.
  • the bounding box generator maps the joint locations from 3D real space coordinates to 2D coordinates in the image frames of respective source images.
  • the bounding box generator 1504 creates bounding boxes for hand joints in image frames in a circular buffer per camera 114.
  • the bounding box is a 128 pixels (width) by 128 pixels (height) portion of the image frame with the hand joint located in the center of the bounding box.
  • the size of the bounding box is 64 pixels x 64 pixels or 32 pixels x 32 pixels.
  • the hand locations of subjects are inferred from locations of elbow and wrist joints. For example, the right hand location of a subject is extrapolated using the location of the right elbow (identified as pi) and the right wrist (identified as p2) as
  • the joints CNN 112a-112n are trained using left and right hand images. Therefore, in such an embodiment, the joints CNN 112a-112n directly identify locations of hand joints in image frames per camera. The hand locations per image frame are used by the bounding box generator 1504 to create a bounding box per identified hand joint.
  • WhatCNN 1506 is a convolutional neural network trained to process the specified bounding boxes in the images to generate a classification of hands of the identified subjects.
  • One trained WhatCNN 1506 processes image frames from one camera.
  • the WhatCNN 1506 identifies whether the hand joint is empty.
  • the WhatCNN 1506 also identifies a SKU (stock keeping unit) number of the inventory item in the hand joint, a confidence value indicating the item in the hand joint is a non-SKU item (i.e. it does not belong to the shopping store inventory) and a context of the hand joint location in the image frame.
  • SKU stock keeping unit
  • the outputs of WhatCNN models 1506 for all cameras 114 are processed by a single WhenCNN model 1508 for a pre-determined window of time.
  • the WhenCNN 1508 performs time series analysis for both hands of subjects to identify whether a subject took a store inventory item from a shelf or put a store inventory item on a shelf.
  • a shopping cart data structure 1510 (also referred to as a log data structure including a list of inventory items) is created per subject to keep a record of the store inventory items in a shopping cart (or basket) associated with the subject.
  • the second image processors subsystem 2604 receives the same data sets comprising subjects identified by joints data structures 800 and corresponding image frames from sequences of image frames per camera as given input to the third image processors.
  • the subsystem 2604 includes background image recognition engines, recognizing semantically significant differences in the background (i.e. inventory display structures like shelves) as they relate to puts and takes of inventory items for example, over time in the images from each camera.
  • a selection logic component (not shown in Fig. 15 A) uses a confidence score to select output from either the second image processors or the third image processors to generate the shopping cart data structure 1510.
  • Fig. 15B shows coordination logic module 1522 combining results of multiple
  • WhatCNN models and giving it as input to a single WhenCNN model As mentioned above, two or more cameras with overlapping fields of view capture images of subjects in real space. Joints of a single subject can appear in image frames of multiple cameras in respective image channel 1520. A separate WhatCNN model identifies SKUs of inventory items in hands (represented by hand joints) of subjects.
  • the coordination logic module 1522 combines the outputs of WhatCNN models into a single consolidated input for the WhenCNN model.
  • the the WhenCNN model 1508 operates on the consolidated input to generate the shopping cart of the subject.
  • FIG. 1617 Detailed implementation of the system comprising multi-CNN pipelines of Fig. 15A is presented in Figs. 16, 17, and 18.
  • the system tracks puts and takes of inventory items by subjects in an area of real space.
  • the area of real space is the shopping store with inventory items placed in shelves organized in aisles as shown in Figs. 2 and 3.
  • shelves containing inventory items can be organized in a variety of different arrangements.
  • shelves can be arranged in a line with their back sides against a wall of the shopping store and the front side facing towards an open area in the real space.
  • a plurality of cameras 114 with overlapping fields of view in the real space produce sequences of images of their corresponding fields of view.
  • the field of view of one camera overlaps with the field of view of at least one other camera as shown in Figs. 2 and 3. Joints CNN - Identification and Update of Subjects
  • Fig. 16 is a flowchart of processing steps performed by joints CNN 112a-l 12n to identify subjects in the real space.
  • the subjects are customers moving in the store in aisles between shelves and other open spaces.
  • the process starts at step 1602.
  • the cameras are calibrated before sequences of images from cameras are processed to identify subjects. Details of camera calibration are presented above.
  • Cameras 1 14 with overlapping fields of view capture images of real space in which subjects are present (step 1604).
  • the cameras are configured to generate synchronized sequences of images.
  • the sequences of images of each camera are stored in respective circular buffers 1502 per camera.
  • a circular buffer (also referred to as a ring buffer) stores the sequences of images in a sliding window of time.
  • a circular buffer stores 1 10 image frames from a corresponding camera.
  • each circular buffer 1502 stores image frames for a time period of 3.5 seconds. It is understood, in other embodiments, the number of image frames (or the time period) can be greater than or less than the example values listed above.
  • Joints CN s 112a-l 12n receive sequences of image frames from corresponding cameras 114 (step 1606).
  • Each joints CNN processes batches of images from a corresponding camera through multiple convolution network layers to identify j oints of subjects in image frames from corresponding camera.
  • Fig. 5 The architecture and processing of images by an example convolutional neural network is presented Fig. 5.
  • the joints of a subject are identified by more than one joints-CNN.
  • the two dimensional (2D) coordinates of joints data structures 600 produced by joints-CNN are mapped to three dimensional (3D) coordinates of the real space to identify joints locations in the real space. Details of this mapping are presented in discussion of Fig. 7 in which the tracking engine 1 10 translates the coordinates of the elements in the arrays of joints data structures corresponding to images in different sequences of images into candidate joints having coordinates in the real space.
  • the joints of a subject are organized in two categories (foot joints and non-foot joints) for grouping the j oints into constellations, as discussed above.
  • the left and right-ankle joint type in the current example are considered foot joints for the purpose of this procedure.
  • heuristics are applied to assign a candidate left foot joint and a candidate right foot joint to a set of candidate joints to create a subject.
  • Other joints from the galaxy of candidate joints can be linked to the subject to build a constellation of some or all of the joint types for the created subject.
  • heuristics are applied to non-foot joints to assign those to the identified subjects.
  • the global metric calculator 702 calculates the global metric value and attempts to minimize the value by checking different combinations of non-foot joints.
  • the global metric is a sum of heuristics organized in four categories as described above.
  • the logic to identify sets of candidate joints comprises heuristic functions based on physical relationships among joints of subjects in real space to identify sets of candidate joints as subjects.
  • the existing subjects are updated using the corresponding non-foot joints. If there are more images for processing (step 1620), steps 1606 to 1618 are repeated, otherwise the process ends at step 1622.
  • a first data sets are produced at the end of the process described above.
  • the first data sets identify subject and the locations of the identified subjects in the real space. In one embodiment, the first data sets are presented above in relation to Fig. 15 A as joints data structures 800 per subject. WhatCN - Classification of Hand Joints
  • Fig. 17 is a flowchart illustrating processing steps to identify inventory items in hands of subjects identified in the real space.
  • the subjects are customers in the shopping store. As the customers move in the aisles and opens spaces, they pick up inventory items stocked in the shelves and put the items in their shopping cart or basket.
  • the image recognition engines identify subjects in the sets of images in the sequences of images received from the plurality of cameras.
  • the system includes the logic to process sets of images in the sequences of images that include the identified subjects to detect takes of inventory items by identified subjects and puts of inventory items on the shelves by identified subjects.
  • the logic to process sets of images includes, for the identified subjects, logic to process images to generate classifications of the images of the identified subjects.
  • the classifications include whether the identified subject is holding an inventory item.
  • the classifications include a first nearness classification indicating a location of a hand of the identified subject relative to a shelf.
  • the classifications include a second nearness classification indicating a location a hand of the identified subject relative to a body of the identified subject.
  • the classifications further include a third nearness classification indicating a location a hand of the identified subject relative to a basket associated with an identified subject.
  • the classifications include an identifier of a likely inventory item.
  • the logic to process sets of images includes, for the identified subjects, logic to identify bounding boxes of data representing hands in images in the sets of images of the identified subjects.
  • the data in the bounding boxes is processed to generate classifications of data within the bounding boxes for the identified subjects.
  • the classifications include whether the identified subject is holding an inventory item.
  • the classifications include a first nearness classification indicating a location of a hand of the identified subject relative to a shelf.
  • the classifications include a second nearness classification indicating a location of a hand of the identified subject relative to a body of the identified subject.
  • the classifications include a third nearness classification indicating a location of a hand of the identified subject relative to a basket associated with an identified subject.
  • the classifications include an identifier of a likely inventory item.
  • the process starts at step 1702.
  • locations of hands (represented by hand joints) of subjects in image frames are identified.
  • the bounding box generator 1504 identifies hand locations of subjects per frame from each camera using joint locations identified in the first data sets generated by joints CN s 1 12a-l 12n as described in Fig. 18.
  • the bounding box generator 1504 processes the first data sets to specify bounding boxes which include images of hands of identified multi -joint subjects in images in the sequences of images. Details of bounding box generator are presented above in discussion of Fig. 15 A.
  • a second image recognition engine receives sequences of images from the plurality of cameras and processes the specified bounding boxes in the images to generate a classification of hands of the identified subjects (step 1708).
  • each of the image recognition engines used to classify the subjects based on images of hands comprises a trained convolutional neural network referred to as a WhatCN 1506.
  • WhatCN s are arranged in multi-CNN pipelines as described above in relation to Fig. 15 A.
  • the input to a WhatCNN is a multi-dimensional array BxWxHxC (also referred to as a BxWxHxC tensor).
  • B is the batch size indicating the number of image frames in a batch of images processed by the WhatCNN.
  • “W” and “FT indicate the width and height of the bounding boxes in pixels
  • the foreground mask, forearm mask and upperarm mask are additional and optional input data sources for the WhatCNN in this example, which the CNN can include in the processing to classify information in the RGB image data.
  • the foreground mask can be generated using mixture of Gaussian algorithms, for example.
  • the forearm mask can be a line between the wrist and elbow providing context produced using information in the Joints data structure .
  • the upperarm mask can be a line between the elbow and shoulder produced using information in the Joints data structure.
  • Different values of B, W, H and C parameters can be used in other embodiments.
  • the size of the bounding boxes is larger e.g., 64 pixels (width) by 64 pixels (height) or 128 pixels (width) by 128 pixels (height).
  • Each WhatCNN 1506 processes batches of images to generate classifications of hands of the identified subjects.
  • the classifications include whether the identified subject is holding an inventory item.
  • the classifications include one or more classifications indicating locations of the hands relative to the shelf and relative to the subject, usable to detect puts and takes.
  • a first nearness classification indicates a location of a hand of the identified subject relative to a shelf.
  • the classifications include in this example a second nearness classification indicating a location a hand of the identified subject relative to a body of the identified subject, where a subject may hold an inventory item during shopping.
  • the classifications in this example further include a third nearness classification indicating a location of a hand of the identified subject relative to a basket associated with an identified subject, where a "basket” in this context is a bag, a basket, a cart or other object used by the subject to hold the inventory items during shopping.
  • the classifications include an identifier of a likely inventory item.
  • the final layer of the WhatCNN 1506 produces logits which are raw values of predictions. The logits are represented as floating point values and further processed, as described below, for generating a classification result.
  • the output "L” per image frame is a raw activation from the WhatCNN 1506.
  • Logits “L” are processed at step 1710 to identify inventory item and context.
  • the first “N” logits represent confidence that the subject is holding one of the “N” inventory items.
  • Logits “L” include an additional five (5) logits which are explained below.
  • the first logit represents confidence that the image of the item in hand of the subj ect is not one of the store SKU items (also referred to as non-SKU item).
  • the second logit indicates a confidence whether the subject is holding an item or not.
  • a large positive value indicates that WhatCNN model has a high level of confidence that the subject is holding an item.
  • a large negative value indicates that the model is confident that the subject is not holding any item.
  • a close to zero value of the second logit indicates that WhatCNN model is not confident in predicting whether the subject is holding an item or not.
  • the next three logits represent first, second and third nearness classifications, including a first nearness classification indicating a location of a hand of the identified subject relative to a shelf, a second nearness classification indicating a location of a hand of the identified subject relative to a body of the identified subject, and a third nearness classification indicating a location of a hand of the identified subject relative to a basket associated with an identified subject.
  • the three logits represent context of the hand location with one logit each indicating confidence that the context of the hand is near to a shelf, near to a basket (or a shopping cart), or near to a body of the subject.
  • the WhatCNN is trained using a training dataset containing hand images in the three contexts: near to a shelf, near to a basket (or a shopping cart), and near to a body of a subject.
  • a "nearness" parameter is used by the system to classify the context of the hand.
  • the system determines the distance of a hand of the identified subject to the shelf, basket (or a shopping cart), and body of the subject to classify the context.
  • the output of a WhatCNN is "L" logits comprised of N SKU logits, 1 Non-SKU logit, 1 holding logit, and 3 context logits as described above.
  • the SKU logits (first N logits) and the non-SKU logit (the first logit following the N logits) are processed by a softmax function.
  • the softmax function transforms a i ⁇ -dimensional vector of arbitrary real values to a i ⁇ -dimensional vector of real values in the range [0, 1] that add up to 1.
  • a softmax function calculates the probabilities distribution of the item over N + 1 items.
  • the output values are between 0 and 1 , and the sum of all the probabilities equals one.
  • the softmax function (for multi-class classification) returns the probabilities of each class.
  • the class that has the highest probability is the predicted class (also referred to as target class).
  • the holding logit is processed by a sigmoid function.
  • the sigmoid function takes a real number value as input and produces an output value in the range of 0 to 1.
  • the output of the sigmoid function identifies whether the hand is empty or holding an item.
  • the three context logits are processed by a softmax function to identify the context of the hand joint location.
  • it is checked if there are more images to process. If true, steps 1704-1710 are repeated, otherwise the process ends at step 1714.
  • the system implements logic to perform time sequence analysis over the classifications of subjects to detect takes and puts by the identified subjects based on foreground image processing of the subjects.
  • the time sequence analysis identifies gestures of the subjects and inventory items associated with the gestures represented in the sequences of images.
  • the outputs of WhatCNNs 1506 in the multi-CNN pipelines are given as input to the WhenCNN 1508 which processes these inputs to detect takes and puts by the identified subjects.
  • the system includes logic, responsive to the detected takes and puts, to generate a log data structure including a list of inventory items for each identified subject.
  • the log data structure is also referred to as a shopping cart data structure 1510 per subject.
  • Fig. 18 presents a process implementing the logic to generate a shopping cart data structure per subject.
  • the process starts at step 1802.
  • the input to WhenCNN 1508 is prepared at step 1804.
  • the input to the WhenCNN is a multi-dimensional array BxCxTxCams, where B is the batch size, C is the number of channels, T is the number of frames considered for a window of time, and Cams is the number of cameras 1 14.
  • the batch size "B" is 64 and the value of "T” is 1 10 image frames or the number of image frames in 3.5 seconds of time.
  • a list of 10 logits per hand joint (20 logits for both hands) is produced.
  • the holding and context logits are part of the "L" logits generated by WhatCN 1506 as described above.
  • the above data structure is generated for each hand in an image frame and also includes data about the other hand of the same subject. For example, if data is for the left hand joint of a subject, corresponding values for the right hand are included as "other" logits.
  • the fifth logit (item number 3 in the list above referred to as log_sku) is the log of SKU logit in "L" logits described above.
  • the sixth logit is the log of SKU logit for other hand.
  • a "roll" function generates the same information before and after the current frame. For example, the seventh logit (referred to as roll(log_sku, -30)) is the log of the SKU logit, 30 frames earlier than the current frame.
  • the eighth logit is the log of the SKU logits for the hand, 30 frames later than the current frame.
  • the ninth and tenth data values in the list are similar data for the other hand 30 frames earlier and 30 frames later than the current frame.
  • a similar data structure for the other hand is also generated, resulting in a total of 20 logits per subject per image frame per camera.
  • the 20 hand logits per subject per frame are consolidated from multi-CNN pipelines.
  • the batch of image frames (64) can be imagined as a smaller window of image frames placed in the middle of a larger window of image frame 1 10 with additional image frames for forward and backward search on both sides.
  • the input BxCxTxCams to WhenCNN 1508 is composed of 20 logits for both hands of subjects identified in batch "B” of image frames from all cameras 114 (referred to as “Cams").
  • the consolidated input is given to a single trained convolutional neural network referred to as WhenCNN model 1508.
  • the output of the WhenCNN model comprises of 3 logits, representing confidence in three possible actions of an identified subject: taking an inventory item from a shelf, putting an inventory item back on the shelf, and no action.
  • the three output logits are processed by a softmax function to predict an action performed.
  • the three classification logits are generated at regular intervals for each subject and results are stored per person along with a time stamp. In one embodiment, the three logits are generated every twenty frames per subject. In such an embodiment, at an interval of every 20 image frames per camera, a window of 110 image frames is formed around the current image frame.
  • a time series analysis of these three logits per subject over a period of time is performed (step 1808) to identify gestures corresponding to true events and their time of occurrence.
  • a non-maximum suppression (NMS) algorithm is used for this purpose.
  • NMS non-maximum suppression
  • As one event i.e. put or take of an item by a subject
  • CNN 1508 multiple times (both from the same camera and from multiple cameras)
  • the NMS removes superfluous events for a subject.
  • NMS is a rescoring technique comprising two main tasks: "matching loss” that penalizes superfluous detections and "joint processing" of neighbors to know if there is a better detection close-by.
  • the true events of takes and puts for each subject are further processed by calculating an average of the SKU logits for 30 image frames prior to the image frame with the true event. Finally, the arguments of the maxima (abbreviated arg max or argmax) is used to determine the largest value.
  • the inventory item classified by the argmax value is used to identify the inventory item put or take from the shelf.
  • the inventory item is added to a log of SKUs (also referred to as shopping cart or basket) of respective subjects in step 1810.
  • the process steps 1804 to 1810 are repeated, if there is more classification data (checked at step 1812). Over a period of time, this processing results in updates to the shopping cart or basket of each subject. The process ends at step 1814.
  • Fig. 19 presents an embodiment of the system in which data from scene process
  • the output from the scene process 1415 is a joints dictionary.
  • keys are unique joint identifiers and values are unique subject identifiers with which the joint is associated. If no subject is associated with a joint, then it is not included in the dictionary.
  • Each video process 1411 receives a joints dictionary from the scene process and stores it into a ring buffer that maps frame numbers to the returned dictionary. Using the returned key -value dictionary, the video processes select subsets of the image at each moment in time that are near hands associated with identified subjects. These portions of image frames around handjoints can be referred to as region proposals.
  • a region proposal is the frame image of hand location from one or more cameras with the subject in their corresponding fields of view.
  • a region proposal is generated by every camera in the system. It includes empty hands as well as hands carrying shopping store inventory items and items not belonging to shopping store inventory.
  • Video processes select portions of image frames containing hand joint per moment in time. Similar slices of foreground masks are generated. The above (image portions of hand joints and foreground masks) are concatenated with the joints dictionary (indicating subjects to whom respective handjoints belong) to produce a multi-dimensional array. This output from video processes is given as input to the WhatCN model.
  • the classification results of the WhatCNN model are stored in the region proposal data structures (produced by video processes). All regions for a moment in time are then given back as input to the scene process.
  • the scene process stores the results in a key -value dictionary, where the key is a subject identifier and the value is a key-value dictionary, where the key is a camera identifier and the value is a region's logits.
  • This aggregated data structure is then stored in a ring buffer that maps frame numbers to the aggregated structure for each moment in time.
  • Fig. 20 presents an embodiment of the system in which the WhenCNN 1508 receives output from a scene process following the hand image classifications performed by the WhatCNN models per video process as explained in Fig. 19.
  • Region proposal data structures for a period of time e.g., for one second, are given as input to the scene process.
  • the input includes 30 time periods and corresponding region proposals.
  • the scene process reduces 30 region proposals (per hand) to a single integer representing the inventory item SKU.
  • the output of the scene process is a key-value dictionary in which the key is a subject identifier and the value is the SKU integer.
  • the WhenCNN model 1508 performs a time series analysis to determine the evolution of this dictionary over time. This results in identification of items taken from shelves and put on shelves in the shopping store.
  • the output of the WhenCNN model is a key -value dictionary in which the key is the subject identifier and the value is logits produced by the WhenCNN.
  • a set of heuristics 2002 is used to determine the shopping cart data structure 1510 per subject. The heuristics are applied to the output of the WhenCNN, joint locations of subjects indicated by their respective joints data structures, and planograms. The planograms are precomputed maps of inventory items on shelves.
  • the heuristics 2002 determine, for each take or put, whether the inventory item is put on a shelf or taken from a shelf, whether the inventory item is put in a shopping cart (or a basket) or taken from the shopping cart (or the basket) or whether the inventory item is close to the identified subject's body.
  • Fig. 21 presents an example architecture of WhatCNN model 1506.
  • This example architecture there are a total of 26 convolutional layers. The dimensionality of different layers in terms of their respective width (in pixels), height (in pixels) and number of channels is also presented.
  • the first convolutional layer 21 13 receives input 211 1 and has a width of 64 pixels, height of 64 pixels and has 64 channels (written as 64x64x64).
  • the details of input to the WhatCNN are presented above.
  • the direction of arrows indicates flow of data from one layer to the following layer.
  • the second convolutional layer 21 15 has a dimensionality of 32x32x64.
  • N is the number of SKUs representing "N" unique inventory items for sale in the shopping store.
  • the five additional logits include the first logit representing confidence that item in the image is a non-SKU item, and the second logit representing confidence whether the subject is holding an item.
  • the next three logits represent first, second and third nearness classifications, as described above.
  • the final output of the WhatCNN is shown at 2137.
  • the example architecture uses batch
  • BN convolutional neural network
  • Figs. 22, 23, and 24 are graphical visualizations of different parts of an implementation of WhatCN 1506.
  • the figures are adapted from graphical visualizations of a WhatCN model generated by TensorBoardTM.
  • TensorBoardTM is a suite of visualization tools for inspecting and understanding deep learning models e.g., convolutional neural networks.
  • Fig. 22 shows a high level architecture of the convolutional neural network model that detects a single hand (“single hand” model 2210).
  • WhatCNN model 1506 comprises two such convolutional neural networks for detecting left and right hands, respectively.
  • the architecture includes four blocks referred to as blockO 2216, blockl 2218, block2 2220, and block3 2222.
  • a block is a higher-level abstraction and comprises multiple nodes representing convolutional layers. The blocks are arranged in a sequence from lower to higher such that output from one block is input to a successive block.
  • the architecture also includes a pooling layer 2214 and a convolution layer 2212. In between the blocks, different non-linearities can be used. In the illustrated embodiment, a ReLU non-linearity is used as described above.
  • the input to the single hand model 2210 is a BxWxHxC tensor defined above in description of WhatCNN 1506.
  • B is the batch size
  • C is the number of channels.
  • the output of the single hand model 2210 is combined with a second single hand model and passed to a fully connected network.
  • the output of the single hand model 2210 is compared with ground truth.
  • a prediction error calculated between the output and the ground truth is used to update the weights of convolutional layers.
  • stochastic gradient descent SGD is used for training WhatCNN 1506.
  • FIG. 23 presents further details of the blockO 2216 of the single hand
  • convolutional neural network model of Fig. 22 It comprises four convolutional layers labeled as convO in box 2310, convl 2318, conv2 2320, and conv3 2322. Further details of the
  • convolutional layer convO are presented in the box 2310.
  • the input is processed by a convolutional layer 2312.
  • the output of the convolutional layer is processed by a batch normalization layer 2314.
  • ReLU non-linearity 2316 is applied to the output of the batch normalization layer 2314.
  • the output of the convolutional layer convO is passed to the next layer convl 2318.
  • the output of the final convolutional layer conv3 is processed through an addition operation 2324. This operation sums the output from the layer conv3 2322 to unmodified input coming through a skip connection 2326. It has been shown by He et al.
  • Fig. 24 is an example implementation of a fully connected layer (FC) 2410.
  • FC fully connected layer
  • the input to the FC layer is processed by a reshape operator 2412.
  • the reshape operator changes the shape of the tensor before passing it to a next layer 2420.
  • Reshaping includes flattening the output from the convolutional layers i.e., reshaping the output from a multi-dimensional matrix to a one-dimensional matrix or a vector.
  • the output of the reshape operator 2412 is passed to a matrix multiplication operator labelled as MatMul 2422.
  • the output from the MatMul 2422 is passed to a matrix plus addition operator labelled as xw_plus_b 2424.
  • xw_plus_b 2424 For each input "x”, the operator 2424 multiplies the input by a matrix "w” and a vector "b" to produce the output, "w” is a trainable parameter associated with the input "x” and "b” is another trainable parameter which is called bias or intercept.
  • the output 2426 from the fully connecter layer 2410 is a BxL tensor as explained above in the description of WhatCNN 1506.
  • B is the batch size
  • N is the number of SKUs representing "N” unique inventory items for sale in the shopping store.
  • a training data set of images of hands holding different inventory items in different contexts, as well as empty hands in different contexts is created.
  • human actors hold each unique SKU inventory item in multiple different ways, at different locations of a test environment.
  • the context of their hands range from being close to the actor's body, being close to the store's shelf, and being close to the actor's shopping cart or basket.
  • the actor performs the above actions with an empty hand as well. This procedure is completed for both left and right hands. Multiple actors perform these actions simultaneously in the same test environment to simulate the natural occlusion that occurs in real shopping stores.
  • Cameras 114 takes images of actors performing the above actions. In one embodiment, twenty cameras are used in this process.
  • the joints CNNs 112a-l 12n and the tracking engine 1 10 process the images to identify joints.
  • the bounding box generator 1504 creates bounding boxes of hand regions similar to production or inference. Instead of classifying these hand regions via the WhatCNN 1506, the images are saved to a storage disk. Stored images are reviewed and labelled. An image is assigned three labels: the inventory item SKU, the context, and whether the hand is holding something or not. This process is performed for a large number of images (up to millions of images).
  • the image files are organized according to data collection scenes.
  • the naming convention for image file identifies content and context of the images.
  • Fig. 25 shows an image file name in an example embodiment.
  • a first part of the file name referred to by a numeral 2502 identifies the data collection scene and also includes the timestamp of the image.
  • a second part 2504 of the file name identifies the source camera. In the example shown in Fig. 25, the image is captured by "camera 4".
  • a third part 2506 of the file name identifies the frame number from the source camera. In the illustrated example, the file name indicates it is the 94,600 th image frame from camera 4.
  • a fourth part 2508 of the file name identifies ranges of x and y coordinates region in the source image frame from which this hand region image is taken.
  • the region is defined between x coordinate values from pixel 117 to 370 and y coordinates values from pixels 370 and 498.
  • a fifth part 2510 of the file name identifies the person id of the actor in the scene. In the illustrated example, the person in the scene has an id "3".
  • stochastic gradient descent is used for training WhatCNN 1506.
  • 64 images are randomly selected from the training data and augmented.
  • the purpose of image augmentation is to diversify the training data resulting in better performance of models.
  • the image augmentation includes random flipping of the image, random rotation, random hue shifts, random Gaussian noise, random contrast changes, and random cropping.
  • the amount of augmentation is a hyperparameter and is tuned through hyperparameter search.
  • the augmented images are classified by WhatCNN 1506 during training. The classification is compared with ground truth and coefficients or weights of WhatCNN 1506 are updated by calculating gradient loss function and multiplying the gradient with a learning rate.
  • the above process is repeated many times (e.g., approximately 1000 times) to form an epoch. Between 50 to 200 epochs are performed. During each epoch, the learning rate is slightly decreased following a cosine annealing schedule.
  • Training of WhenCNN 1508 is similar to the training of WhatCNN 1506 described above, using backpropagations to reduce prediction error.
  • Actors perform a variety of actions in the training environment.
  • the training is performed in a shopping store with shelves stocked with inventory items.
  • actions performed by actors include, take an inventory item from a shelf, put an inventory item back on a shelf, put an inventory item into a shopping cart (or a basket), take an inventory item back from the shopping cart, swap an item between left and right hands, put an inventory item into the actor's nook.
  • a nook refers to a location on the actor's body that can hold an inventory item besides the left and right hands.
  • Some examples of nook include, an inventory item squeezed between a forearm and upper arm, squeezed between a forearm and a chest, squeezed between neck and a shoulder.
  • the cameras 114 record videos of all actions described above during training.
  • the videos are reviewed and all image frames are labelled indicating the timestamp and the action performed. These labels are referred to as action labels for respective image frames.
  • the image frames are processed through the multi-CNN pipelines up to the WhatCNNs 1506 as described above for production or inference.
  • the output of WhatCNNs along withthe associated action labels are then used to train the WhenCNN 1508, with the action labels acting as ground truth.
  • Stochastic gradient descent (SGD) with a cosine annealing schedule is used for training as described above for training of WhatCNN 1506.
  • temporal augmentation is also applied to image frames during training of the WhenCNN.
  • Some examples include mirroring, adding Gaussian noise, swapping the logits associated with left and right hands, shortening the time, shortening the time series by dropping image frames, lengthening the time series by duplicating frames, and dropping the data points in the time series to simulate spottiness in the underlying model generating input for the WhenCNN.
  • Mirroring includes reversing the time series and respective labels, for example a put action becomes a take action when reversed.
  • Fig. 26 presents a high level schematic of a system in accordance with an implementation. Because Fig. 26 is an architectural diagram, certain details are omitted to improve the clarity of description.
  • the system presented in Fig. 26 receives image frames from a plurality of cameras 114.
  • the cameras 1 14 can be synchronized in time with each other, so that images are captured at the same time, or close in time, and at the same image capture rate. Images captured in all the cameras covering an area of real space at the same time, or close in time, are synchronized in the sense that the synchronized images can be identified in the processing engines as representing different views at a moment in time of subjects having fixed positions in the real space.
  • the cameras 1 14 are installed in a shopping store (such as a supermarket) such that sets of cameras (two or more) with overlapping fields of view are positioned over each aisle to capture images of real space in the store. There are "n" cameras in the real space. Each camera produces a sequence of images of real space corresponding to its respective field of view.
  • a subject identification subsystem 2602 processes image frames received from cameras 1 14 to identify and track subjects in the real space.
  • the first image processors include subject image recognition engines.
  • the subject image recognition engines receive corresponding sequences of images from the plurality of cameras, and process images to identify subjects represented in the images in the corresponding sequences of images.
  • the system includes per camera image recognition engines as described above for identifying and tracking multi-joint subjects.
  • Alternative image recognition engines can be used, including examples in which only one "joint" is recognized and tracked per individual, or other features or other types of images data over space and time are utilized to recognize and track subjects in the real space being processed.
  • a "semantic diffing" subsystem 2604 includes background image recognition engines, receiving corresponding sequences of images from the plurality of cameras and recognize semantically significant differences in the background (i.e. inventory display structures like shelves) as they relate to puts and takes of inventory items for example, over time in the images from each camera.
  • the second image processors receive output of the subject identification subsystem 2602 and image frames from cameras 114 as input.
  • the second image processors mask the identified subjects in the foreground to generate masked images.
  • the masked images are generated by replacing bounding boxes that correspond with foreground subjects with background image data.
  • the background image recognition engines process the masked images to identify and classify background changes represented in the images in the corresponding sequences of images.
  • the background image recognition engines comprise convolutional neural networks.
  • the second image processors process identified background changes to make a first set of detections of takes of inventory items by identified subjects and of puts of inventory items on inventory display structures by identified subjects.
  • the first set of detections are also referred to as background detections of puts and takes of inventory items.
  • the first detections identify inventory items taken from the shelves or put on the shelves by customers or employees of the store.
  • the semantic diffing subsystem includes the logic to associate identified background changes with identified subjects.
  • a region proposals subsystem 2606 include foreground image recognition engines, receiving corresponding sequences of images from the plurality of cameras 114, and recognize semantically significant objects in the foreground (i.e. shoppers, their hands and inventory items) as they relate to puts and takes of inventory items for example, over time in the images from each camera.
  • the subsystem 2606 also receives output of the subject identification subsystem 2602.
  • the third image processors process sequences of images from cameras 114 to identify and classify foreground changes represented in the images in the corresponding sequences of images.
  • the third image processors process identified foreground changes to make a second set of detections of takes of inventory items by identified subjects and of puts of inventory items on inventory display structures by identified subjects.
  • the second set of detections are also referred to as foreground detection of puts and takes of inventory items. In the example of a shopping store, the second set of detections identify takes of inventory items and puts of inventory items on inventory display structures by customers and employees of the store.
  • the system described in Fig. 26 includes a selection logic component 2608 to process the first and second sets of detections to generate log data structures including lists of inventory items for identified subjects. For a take or put in the real space, the selection logic 2608 selects the output from either the semantic diffing subsystem 2604 or the region proposals subsystem 2606. In one embodiment, the selection logic 2608 uses a confidence score generated by the semantic diffing subsystem for the first set of detections and a confidence score generated by the region proposals subsystem for a second set of detections to make the selection. The output of the subsystem with a higher confidence score for a particular detection is selected and used to generate a log data structure 1510 (also referred to as a shopping cart data structure) including a list of inventory items associated with identified foreground subjects.
  • a log data structure 1510 also referred to as a shopping cart data structure
  • Fig. 27 presents subsystem components implementing the system for tracking changes by subjects in an area of real space.
  • the system comprises of the plurality of cameras 1 14 producing respective sequences of images of corresponding fields of view in the real space.
  • the field of view of each camera overlaps with the field of view of at least one other camera in the plurality of cameras as described above.
  • the sequences of image frames corresponding to the images produced by the plurality of cameras 1 14 are stored in a circular buffer 1502 (also referred to as a ring buffer) per camera 1 14.
  • Each image frame has a timestamp, identity of the camera (abbreviated as "camera id”), and a frame identity
  • Circular buffers 1502 store a set of consecutively timestamped image frames from respective cameras 1 14.
  • the cameras 114 are configured to generate synchronized sequences of images.
  • the subject identification subsystem 2602 (also referred to as the first image processors), includes subject image recognition engines, receiving corresponding sequences of images from the plurality of cameras 114.
  • the subject image recognition engines process images to identify subjects represented in the images in the corresponding sequences of images.
  • the subject image recognition engines are implemented as convolutional neural networks (CN s) referred to as joints CNN 112a-112n.
  • the outputs of joints CNNs 112a-112n corresponding to cameras with overlapping fields of view are combined to map the location of joints from 2D image coordinates of each camera to 3D coordinates of real space.
  • a background image store 2704 in the semantic diffing subsystem 2604, stores masked images (also referred to as background images in which foreground subjects have been removed by masking) for corresponding sequences of images from cameras 1 14.
  • the background image store 2704 is also referred to as a background buffer.
  • the size of the masked images is the same as the size of image frames in the circular buffer 1502.
  • a masked image is stored in the background image store 2704 corresponding to each image frame in the sequences of image frames per camera.
  • the semantic diffing subsystem 2604 (or the second image processors) includes a mask generator 2724 producing masks of foreground subjects represented in the images in the corresponding sequences of images from a camera.
  • a mask generator 2724 produces masks of foreground subjects represented in the images in the corresponding sequences of images from a camera.
  • one mask generator processes sequences of images per camera.
  • the foreground subjects are customers or employees of the store in front of the background shelves containing items for sale.
  • the joint data structures 800 and image frames from the circular buffer 1502 are given as input to the mask generator 2724.
  • the joint data structures identify locations of foreground subjects in each image frame.
  • the mask generator 2724 generates a bounding box per foreground subject identified in the image frame.
  • the mask generator 2724 uses the values of the x and y coordinates of joint locations in 2D image frame to determine the four boundaries of the bounding box.
  • a minimum value of x (from all x values of joints for a subject) defines the left vertical boundary of the bounding box for the subject.
  • a minimum value of y defines the bottom horizontal boundary of the bounding box.
  • the mask generator 2724 produces bounding boxes for foreground subjects using a convolutional neural network-based person detection and localization algorithm. In such an embodiment, the mask generator 2724 does not use the joint data structures 800 to generate bounding boxes for foreground subjects.
  • the semantic diffing subsystem 2604 (or the second image processors) include a mask logic to process images in the sequences of images to replace foreground image data representing the identified subjects with background image data from the background images for the corresponding sequences of images to provide the masked images, resulting in a new background image for processing.
  • the mask logic processes images in the sequences of images to replace foreground image data defined by the image masks with background image data.
  • the background image data is taken from the background images for the corresponding sequences of images to generate the corresponding masked images.
  • a background image in the background image store 2704 is the same as its corresponding image frame in the sequences of images per camera.
  • the mask generator 2724 creates a bounding box of the customer and sends it to a mask logic component 2702.
  • the mask logic component 2702 combines, such as by averaging or summing by pixel, sets of N masked images in the sequences of images to generate sequences of factored images for each camera.
  • the second image processors identify and classify background changes by processing the sequence of factored images.
  • a factored image can be generated, for example, by taking an average value for pixels in the N masked images in the sequence of masked images per camera.
  • the value of N is equal to the frame rate of cameras 1 14, for example if the frame rate is 30 FPS (frames per second), the value of N is 30.
  • the masked images for a time period of one second are combined to generate a factored image. Taking the average pixel values minimizes the pixel fluctuations due to sensor noise and luminosity changes in the area of real space.
  • the second image processors identify and classify background changes by processing the sequence of factored images.
  • a factored image in the sequences of factored images is compared with the preceding factored image for the same camera by a bit mask calculator 2710. Pairs of factored images 2706 are given as input to the bit mask calculator 2710 to generate a bit mask identifying changes in corresponding pixels of the two factored images.
  • the bit mask has Is at the pixel locations where the difference between the corresponding pixels' (current and previous factored image) RGB (red, green and blue channels) values is greater than a "difference threshold".
  • the value of the difference threshold is adjustable. In one embodiment, the value of the difference threshold is set at 0.1.
  • the bit mask and the pair of factored images (current and previous) from sequences of factored images per camera are given as input to background image recognition engines.
  • the background image recognition engines comprise convolutional neural networks and are referred to as ChangeCN 2714a-2714n.
  • a single ChangeCN processes sequences of factored images per camera.
  • the masked images from corresponding sequences of images are not combined.
  • the bit mask is calculated from the pairs of masked images.
  • the pairs of masked images and the bit mask is then given as input to the ChangeCNN.
  • the input to a ChangeCNN model in this example consists of seven (7) channels including three image channels (red, green and blue) per factored image and one channel for the bit mask.
  • the ChangeCNN comprises of multiple convolutional layers and one or more fully connected (FC) layers.
  • the ChangeCNN comprises of the same number of convolutional and FC layers as the JointsCNN 112a- 112n as illustrated in Fig. 5.
  • the background image recognition engines identify and classify changes in the factored images and produce change data structures for the corresponding sequences of images.
  • the change data structures include coordinates in the masked images of identified background changes, identifiers of an inventory item subject of the identified background changes and classifications of the identified background changes.
  • the classifications of the identified background changes in the change data structures classify whether the identified inventory item has been added or removed relative to the background image.
  • the ChangeCNN As multiple items can be taken or put on the shelf simultaneously by one or more subjects, the ChangeCNN generates a number "B" overlapping bounding box predictions per output location.
  • a bounding box prediction corresponds to a change in the factored image.
  • the shopping store has a number "C” unique inventory items, each identified by a unique SKU.
  • the ChangeCNN predicts the SKU of the inventory item subject of the change.
  • the ChangeCNN identifies the change (or inventory event type) for every location (pixel) in the output indicating whether the item identified is taken from the shelf or put on the shelf.
  • the above three parts of the output from ChangeCNN are described by an expression "5 * B + C + 1".
  • Each bounding box "B” prediction comprises of five (5) numbers, therefore "B” is multiplied by 5. These five numbers represent the "x" and "y” coordinates of the center of the bounding box, the width and height of the bounding box.
  • the fifth number represents
  • the value of "B" equals 4.
  • ChangeCNN is then expressed as "W * H * (5 * B + C + 1)".
  • the bounding box output model is based on object detection system proposed by Redmon and Farhadi in their paper, "YOLO9000: Better, Faster, Stronger” published on December 25, 2016. The paper is available at
  • the outputs of ChangeCNN 2714a-2714n corresponding to sequences of images from cameras with overlapping fields of view are combined by a coordination logic component 2718.
  • the coordination logic component processes change data structures from sets of cameras having overlapping fields of view to locate the identified background changes in real space.
  • the coordination logic component 2718 selects bounding boxes representing the inventory items having the same SKU and the same inventory event type (take or put) from multiple cameras with overlapping fields of view.
  • the selected bounding boxes are then triangulated in the 3D real space using triangulation techniques described above to identify the location of the inventory item in 3D real space. Locations of shelves in the real space are compared with the triangulated locations of the inventory items in the 3D real space. False positive predictions are discarded.
  • Triangulated locations of bounding boxes in the 3D real space that map to a shelf are considered true predictions of inventory events.
  • the classifications of identified background changes in the change data structures produced by the second image processors classify whether the identified inventory item has been added or removed relative to the background image.
  • the classifications of identified background changes in the change data structures indicate whether the identified inventory item has been added or removed relative to the background image and the system includes logic to associate background changes with identified subjects. The system makes detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects.
  • a log generator 2720 implements the logic to associate changes identified by true predictions of changes with identified subjects near the location of the change. In an embodiment utilizing the joints identification engine to identify subjects, the log generator 2720 determines the positions of hand joints of subjects in the 3D real space using joint data structures 800. A subject whose hand joint location is within a threshold distance to the location of a change at the time of the change is identified. The log generator associates the change with the identified subject. [0238] In one embodiment, as described above, N masked images are combined to generate factored images which are then given as input to the ChangeCN . Consider, N equals the frame rate (frames per second) of the cameras 1 14.
  • the positions of hands of subjects during a one second time period are compared with the location of the change to associate the changes with identified subjects. If more than one subject's hand joint locations are within the threshold distance to a location of a change, then association of the change with a subj ect is deferred to output of the foreground image processing subsystem 2606.
  • the foreground image processing (region proposals) subsystem 2606 (also referred to as the third image processors) include foreground image recognition engines receiving images from the sequences of images from the plurality of cameras.
  • the third image processors include logic to identify and classify foreground changes represented in the images in the corresponding sequences of images.
  • the region proposals subsystem 2606 produces a second set of detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects.
  • the subsystem 2606 includes the bounding box generator 1504, the WhatCN 1506 and the
  • the system described in Fig. 27 includes the selection logic to process the first and second sets of detections to generate log data structures including lists of inventory items for identified subjects.
  • the first set of detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects are generated by the log generator 2720.
  • the first set of detections are determined using the outputs of second image processors and the joint data structures 800 as described above.
  • the second set of detections of takes of inventory items by the identified subjects and of puts of inventory items on inventory display structures by the identified subjects are determined using the output of the third image processors.
  • the selection logic controller 2608 selects the output from either the second image processors (semantic diffing subsystem 2604) or the third image processors (region proposals subsystem 2606). In one embodiment, the selection logic selects the output from an image processor with a higher confidence score for prediction of that inventory event.
  • Figs. 28A and 28B present detailed steps performed by the semantic diffing subsystem 2604 to track changes by subj ects in an area of real space.
  • the subjects are customers and employees of the store moving in the store in aisles between shelves and other open spaces.
  • the process starts at step 2802.
  • the cameras 114 are calibrated before sequences of images from cameras are processed to identify subjects. Details of camera calibration are presented above. Cameras 114 with overlapping fields of view capture images of real space in which subjects are present.
  • the cameras are configured to generate synchronized sequences of images at the rate of N frames per second.
  • the sequences of images of each camera are stored in respective circular buffers 1502 per camera at step 2804.
  • a circular buffer also referred to as a ring buffer
  • the background image store 2704 is initialized with initial image frame in the sequence of image frames per camera with no foreground subjects (step 2806).
  • bounding boxes per subject are generated using their corresponding joint data structures 800 as described above (step 2808).
  • a masked image is created by replacing the pixels in the bounding boxes per image frame by pixels at the same locations from the background image from the background image store 2704.
  • the masked image corresponding to each image in the sequences of images per camera is stored in the background image store 2704.
  • the z ' th masked image is used as a background image for replacing pixels in the following (z ' +l) image frame in the sequence of image frames per camera.
  • N masked images are combined to generate factored images.
  • a difference heat map is generated by comparing pixel values of pairs of factored images.
  • the difference between pixels at a location (x, y) in a 2D space of the two factored images (fll and fi2) is calculated as shown below in equation 1 :
  • the difference between the pixels at the same x and y locations in the 2D space is determined using the respective intensity values of red, green and blue (RGB) channels as shown in the equation.
  • the above equation gives a magnitude of the difference (also referred to as Euclidean norm) between corresponding pixels in the two factored images.
  • the difference heat map can contain noise due to sensor noise and luminosity changes in the area of real space.
  • a bit mask is generated for a difference heat map. Semantically meaningful changes are identified by clusters of I s (ones) in the bit mask. These clusters correspond to changes identifying inventory items taken from the shelf or put on the shelf. However, noise in the difference heat map can introduce random Is in the bit mask. Additionally, multiple changes (multiple items take from or put on the shelf) can introduce overlapping clusters of I s.
  • image morphology operations are applied to the bit mask. The image morphology operations remove noise (unwanted Is) and also attempt to separate overlapping clusters of I s. This results in a cleaner bit mask comprising clusters of I s corresponding to semantically meaningful changes.
  • a kernel consists of Is arranged in a rectangular matrix in a variety of sizes. Kernels of different shapes (for example, circular, elliptical or cross-shaped) are created by adding 0's at specific locations in the matrix. Kernels of different shapes are used in image morphology operations to achieve desired results in cleaning bit masks.
  • erosion operation a kernel slides (or moves) over the bit mask. A pixel (either 1 or 0) in the bit mask is considered 1 if all the pixels under the kernel are I s. Otherwise, it is eroded (changed to 0). Erosion operation is useful in removing isolated I s in the bit mask. However, erosion also shrinks the clusters of Is by eroding the edges.
  • Dilation operation is the opposite of erosion. In this operation, when a kernel slides over the bit mask, the values of all pixels in the bit mask area overlapped by the kernel are changed to 1 , if value of at least one pixel under the kernel is 1. Dilation is applied to the bit mask after erosion to increase the size clusters of Is. As the noise is removed in erosion, dilation does not introduce random noise to the bit mask. A combination of erosion and dilation operations are applied to achieve cleaner bit masks. For example the following line of computer program code applies a 3x3 filter of Is to the bit mask to perform an "open" operation which applies erosion operation followed by dilation operation to remove noise and restore the size of clusters of I s in the bit mask as described above. The above computer program code uses OpenCV (open source computer vision) library of programming functions for real time computer vision applications. The library is available at https://opencv.org/.
  • OpenCV open source computer vision
  • bit mask cv2.morphologyEx(bit_mask, cv2.MORPH OPEN, self.kernel_3x3,
  • a "close” operation applies dilation operation followed by erosion operation. It is useful in closing small holes inside the clusters of Is.
  • the following program code applies a close operation to the bit mask using a 30x30 cross-shaped filter,
  • the bit mask and the two factored images (before and after) are given as input to a convolutional neural network (referred to as ChangeCN above) per camera.
  • the outputs of ChangeCNN are the change data structures.
  • outputs from ChangeCN s with overlapping fields of view are combined using triangulation techniques described earlier.
  • a location of the change in the 3D real space is matched with locations of shelves. If location of an inventory event maps to a location on a shelf, the change is considered a true event (step 2824). Otherwise, the change is a false positive and is discarded.
  • True events are associated with a foreground subject.
  • the foreground subject is identified.
  • the joints data structure 800 is used to determine location of a hand joint within a threshold distance of the change. If a foreground subject is identified at the step 2828, the change is associated to the identified subject at a step 2830. If no foreground subject is identified at the step 2828, for example, due to multiple subjects' hand joint locations within the threshold distance of the change. Then redundant detection of the change by region proposals subsystem is selected at a step 2832. The process ends at a step 2834.
  • a training data set of seven channel inputs is created to train the ChangeCNN.
  • the SKU number for the inventory item in the change is identified and included in the label for the image along with the bounding box.
  • An event type identifying take or put of inventory item is also included in the label of the bounding box.
  • the label for each bounding box identifies, its location on the factored image, the SKU of the item and the event type.
  • a factored image can have more than one bounding boxes. The above process is repeated for every change in all collected factored images in the training data set. A pair of factored images along with the bit mask forms a seven channel input to the ChangeCN .
  • the ChangeCNN identify and classify background changes represented in the factored images in the corresponding sequences of images in the training data set.
  • the ChangeCNN process identified background changes to make a first set of detections of takes of inventory items by identified subjects and of puts of inventory items on inventory display structures by identified subj ects.
  • the output of the ChangeCNN is compared with the ground truth as indicated in labels of training data set.
  • a gradient for one or more cost functions is calculated.
  • the gradient(s) are then propagated to the convolutional neural network (CNN) and the fully connected (FC) neural network so that the prediction error is reduced causing the output to be closer to the ground truth.
  • CNN convolutional neural network
  • FC fully connected
  • a softmax function and a cross-entropy loss function is used for training of the ChangeCNN for class prediction part of the output.
  • the class prediction part of the output includes an SKU identifier of the inventory item and the event type i.e., a take or a put.
  • a second loss function is used to train the ChangeCNN for prediction of bounding boxes.
  • This loss function calculates intersection over union (IOU) between the predicted box and the ground truth box. Area of intersection of bounding box predicted by the ChangeCNN with the true bounding box label is divided by the area of the union of the same bounding boxes. The value of IOU is high if the overlap between the predicted box and the ground truth boxes is large. If more than one predicted bounding boxes overlap the ground truth bounding box, then the one with highest IOU value is selected to calculate the loss function. Details of the loss function are presented by Redmon et. al, in their paper, "You Only Look Once: Unified, Real-Time Object Detection” published on May 9, 2016. The paper is available at
  • the system for tracking puts and takes of inventory items by subj ects in an area of real space described above also includes one or more of the following features.
  • a region proposal is the frame image of hand location from all different cameras covering the person.
  • a region proposal is generated by every camera in the system. It includes empty hands as well as hands carrying store items.
  • the WhatCN model is the WhatCN model.
  • a region proposal can be used as input to image classification using a deep learning algorithm.
  • This classification engine is called a "WhatCNN" model. It is an in-hand classification model. It classifies the things that are in hands. In-hand image classification can operate even though parts of the obj ect are occluded by the hand. Smaller items may be occluded up to 90% by the hand.
  • the region for image analysis by the WhatCNN model is intentionally kept small in some embodiments because it is computationally expensive.
  • Each camera can have a dedicated GPU. This is performed for every hand image from every camera for every frame.
  • a confidence weight is also assigned to that image (one camera, one point in time).
  • the classification algorithm outputs logits over the entire list of stock keeping units (SKUs) to produce a product and service identification code list of the store for n items and one additional for an empty hand (n+1).
  • Each video process receives the key-value dictionary from the scene process and stores it into a ring buffer that maps frame numbers to the returned dictionary.
  • the video selects subsets of the image at each moment in time that are near hands associated with known people. These regions are numpy slices. We also take a similar slice around foreground masks and the raw output feature arrays of the Joints CNN. These combined regions are concatenated together into a single multidimensional numpy array and stored in a data structure that holds the numpy array as well as the person ID with which the region is associated and which hand from the person the region came from.
  • a CNN dedicated to classification referred to as a WhatCNN.
  • the output of this CNN is a flat array of floats of size N+1 , where N is the number of unique SKUs in the store, and the final class represents the nil class, or empty hand.
  • the floats in this array are referred to as logits.
  • the scene process receives all regions from all videos at a moment in time and stores the results in a key-value dictionary, where the key is a person ID and the value is a key- value dictionary, where the key is a camera ID and the value is a region's logits.
  • This aggregated data structure is then stored in a ring buffer that maps frame numbers to the aggregated structure for each moment in time.
  • the images from different cameras processed by the WhatCNN model are combined over a period of time (multiple cameras over a period of time).
  • An additional input to this model is hand location in 3D space, triangulated from multiple cameras.
  • Another input to this algorithm is the distance of a hand from a planogram of the store. In some embodiments, the planogram can be used to identify if the hand is close to a shelf containing a particular item (e.g. cheerios boxes).
  • Another input to this algorithm is the foot location on the store.
  • the second classification model uses time series analysis to determine whether the object was picked up from the shelf or placed on the shelf. The images are analyzed over a period of time to make the determination of whether the object that was in the hand in earlier image frames has been put back in the shelf or has been picked up from the shelf.
  • This model also includes output from the shelf model as its input to identify what object this person has picked.
  • the scene process waits for 30 or more aggregated structures to accumulate, representing at least a second of real time, and then performs a further analysis to reduce the aggregated structure down to a single integer for each person ID-hand pair, where the integer is a unique ID representing a SKU in the store. For a moment in time this information is stored in a key-value dictionary where keys are person ID-hand pairs, and values are the SKU integer. This dictionary is stored over time in a ring buffer that maps frame numbers to each dictionary for that moment in time.
  • the output of the WhenCNN is stored in a ring buffer that maps frame numbers to a key-value dictionary where keys are person IDs and values are the extended logits emitted by the WhenCNN.
  • the system can send the inventory list to the shopper's phone.
  • the phone displays the user's inventory and asks for confirmation to charge their stored credit card information. If the user accepts, their credit card will be charged. If they do not have a credit card known in the system, they will be asked to provide credit card information.
  • the shopper may also approach an in-store kiosk.
  • the system identifies when the shopper is near the kiosk and will send a message to the kiosk to display the inventory of the shopper.
  • the kiosk asks the shopper to accept the charges for the inventory. If the shopper accepts, they may then swipe their credit card or insert cash to pay.
  • Fig. 16 presents an illustration of the WhenCNN model for region proposals.
  • This feature identifies misplaced items when they are placed back by a person on a random shelf. This causes problems in object identification because the foot and hand location with respect to the planogram will be incorrect. Therefore, the system builds up a modified planogram over time. Based on prior time series analysis, the system is able to determine if a person has placed an item back in the shelf. Next time, when an object is picked up from that shelf location, the system knows that there is at least one misplaced item in that hand location. Correspondingly, the algorithm will have some confidence that the person can pick up the misplaced item from that shelf. If the misplaced item is picked up from the shelf, the system subtracts that item from that location and therefore, the shelf does not have that item anymore. The system can also inform a clerk about a misplaced item via an app so that the clerk can move that item to its correct shelf.
  • An alternative technology for background image processing comprises a background subtraction algorithm to identify changes to items (items removed or placed) on the shelves. This is based on changes at the pixel level. If there are persons in front of the shelf, then the algorithm stops so that it does not take into account pixel changes due to presence of persons. Background subtraction is a noisy process. Therefore, a cross-camera analysis is conducted. If enough cameras agree that there is a "semantically meaningful" change in the shelf, then the system records that there is a change in that part of the shelf.
  • the next step is to identify whether that change is a "put" or a "get” change.
  • the time series analysis of the second classification model is used.
  • a region proposal for that particular part of the shelf is generated and passed through the deep learning algorithm. This is easier than in-hand image analysis because the object is not occluded inside a hand.
  • a fourth input is given to the algorithm in addition to the three typical RGB inputs.
  • the fourth channel is the background information.
  • the output of the shelf or semantic diffing is input again to the second classification model (time-series analysis model).
  • Semantic diffing in this approach includes the following steps:
  • Images from a camera are compared to earlier images from the same camera.
  • Each corresponding pixel between the two images is compared via a Euclidean distance in RGB space.
  • a collection of image morphology filters are used to remove noise from the marked image. 5. We then search for large collections of marked pixels and form bounding boxes around them.
  • Predefined shopping scripts are used to collect good quality data of images. These images are used for training of algorithms.
  • Data collection includes the following steps:
  • a script is automatically generated telling a human actor what actions to take.
  • actions are randomly sampled from a collection of actions including: take item X, place item X, hold item X for Y seconds.
  • the script serves as an input label to machine leaming models (such as the CNNs) that train on the videos of actors. 6.
  • the Store App has several main capabilities; providing data analytic
  • Data is collected by the platform and can be used in a variety of ways.
  • the derivative data is used to perform various kinds of analytics on stores, the shopping experiences they provide, and customer interactions with products, environment, and other people.
  • the data is stored and used in the background to perform analyses of the store and customer interactions.
  • the Store App will display some of the visualizations of this data to retailers. Other data is stored and queried when the data point is desired.
  • the platform visualizes a retailer's floor plan, shelf layouts, and other store environments with overlays showing levels of various kinds of activity.
  • the platform tracks all of a store's SKUs. When an item gets put in the incorrect place, the platform will know where that item is and build a log. At some threshold, or immediately, store employees may be alerted to the misplaced item. Alternatively, the staff may access the
  • Misplaced Item Map in the Store App When convenient, staff can then quickly locate and correct misplaced items.
  • Granular breakdown of product interactions such as:
  • impressions will be decided.
  • the threshold is met, the products would make it to her list and be sent to her soon after leaving the store.
  • the shopper could be sent an email a period of time later that offered product(s) on sale or other special information. These products will be items they expressed interest in, but did not purchase.
  • the Shopper App automatically checks people out when they exit the store.
  • the platform does not require shoppers to have or use the Shopper App to use the store.
  • the shopper will have the opportunity to review what is displayed. If they agree with the information on the display they can either enter cash into the machine (if that capability is built into the hardware (e.g. self-checkout machines)) or they swipe their credit or debit card. They can then exit the store.
  • the Shopper App Through use of an app, the Shopper App, the customer can exit the store with merchandise and automatically be charged and given a digital receipt.
  • the shopper must open their app at any time while within the store's shopping area.
  • the platform will recognize a unique image that is displayed on the shopper's device.
  • the platform will tie them to their account (Customer Association), and regardless if they keep the app open or not, will be able to remember who they are throughout their time in the store's shopping area.
  • the Shopper App will display the items in shopper's Shopping Cart. If the shopper wishes, they can view product information about each item they pick up (i.e. gets added to their shopping cart). Product information is stored either with the store's systems or added to a platform. The ability for updating that information, such as offering product sales or displaying prices, is an option the retailer can request/purchase or develop.
  • the Platform will maintain the shopper's Shopping Cart and correctly charge them once they exit the store.
  • the Shopper App also has mapping information on its development roadmap. It can tell a customer where to find items in the store if the customer requests the information by typing in the item being sought. At a later date, we will take a shopper's shopping list (entered into the app manually or through other intelligent systems) and display the fastest route through the store to collect all the desired items. Other filters, such as 'Bagging Preference' may be added. The Bagging Preference filter allows a shopper to not follow the fastest route, but to gather sturdier items first, then more fragile items later.
  • the system assigns color codes (green and yellow) to the customers walking in the store based on the confidence level.
  • the green color coded customers are either logged into the system or the system has a high confidence about them.
  • Yellow color coded customers have one or more items that are not predicted with high confidence.
  • a clerk can look at the yellow dots and click on them to identify problem items, walk up to the customer and fix the problem.
  • a host of analytics information is gathered about the customer such as how much time a customer spent in front of a particular shelf. Additionally, the system tracks the location where a customer is looking (impression on the system), and the items which a customer picked and put back on the shelf. Such analytics are currently available in ecommerce but not available in retail stores.
  • System to generate customer shopping analytics including location-based impressions, directional impressions, A/B analysis, customer recognition, group dynamics etc.
  • the technology described herein can support Cashier-free Checkout. Go to Store. Take Things. Leave.
  • Cashier-free Checkout is a pure machine vision and deep learning based system.
  • Techniques & Capabilities include the following:
  • Ceiling cameras must be installed with double coverage of all areas of the store. Requires between 2 and 6 cameras for a typical aisle.
  • An on-premise GPU cluster can fit into one or two server racks in a back office.
  • Example systems can be integrated with or include Point of Sale and Inventory
  • a first system, method and computer program product for capturing arrays of images in stores using synchronized cameras is a first system, method and computer program product for capturing arrays of images in stores using synchronized cameras.
  • a second system, method and computer program product to identify joints in images, and sets of joints of individual persons.
  • An eighth system, method and computer program product to generate an inventory array per person using region proposals and get put analysis e.g. Outputs of
  • a ninth system, method and computer program product to identify, track and update locations of misplaced items on shelves.
  • a fourteenth system, method and computer program product to perform checkout and collect payment from member customers.
  • a seventeenth system, method and computer program product to track customers using for example color codes to help clerks identify incorrectly identified items in a customer's cart.
  • An eighteenth system, method and computer program product to generate customer shopping analytics including one or more of location-based impressions, directional impressions, A/B analysis, customer recognition, group dynamics etc.
  • Described herein is a method for tracking puts and takes of inventory items by subjects in an area of real space, comprising:
  • processing the first data sets to specify bounding boxes which include images of hands of identified subjects in images in the sequences of images;
  • the first data sets can comprise for each identified subject sets of candidate joints having coordinates in real space.
  • This described method can include processing the first data sets to specify bounding boxes includes specifying bounding boxes based on locations of joints in the sets of candidate joints for each subject.
  • one or both of the first and the second image recognition engines can comprise convolutional neural networks.
  • This described method can include processing the classifications of bounding boxes using convolutional neural networks.
  • a computer program product and products are described which include a computer readable memory comprising a non-transitory data storage medium, and computer instructions stored in the memory executable by a computer to track puts and takes of inventory items by subjects in an area of real space by any of the herein described processes.
  • a system comprising a plurality of cameras producing a sequences of images including a hand of a subject; and a processing system coupled to the plurality of cameras, the processing system including a hand image recognition engine, receiving the sequence of images, to generate classifications of the hand in time sequence, and logic to process the classifications of the hand from the sequence of images to identify an action by the subject, wherein, the action is one of puts and takes of inventory items.
  • the system can include logic to identify locations of joints of the subject in the images in the sequences of images, and to identify bounding boxes in corresponding images that include the hands of the subject based on the identified joints.
  • a computer program listing appendix accompanies the specification, and includes portions of an example of a computer program to implement certain parts of the system provided in this application.
  • the appendix includes examples of heuristics to identify joints of subjects and inventory items.
  • the appendix presents computer program code to update a subject's shopping cart data structure.
  • the appendix also includes a computer program routine to calculate learning rate during training of a convolutional neural network.
  • the appendix includes a computer program routine to store classification results of hands of subjects from a convolutional neural network in a data structure per hand per subject per image frame from each camera.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Accounting & Taxation (AREA)
  • Evolutionary Computation (AREA)
  • Finance (AREA)
  • Strategic Management (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • General Engineering & Computer Science (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Medical Informatics (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Human Resources & Organizations (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)

Abstract

L'invention concerne des systèmes et des techniques pour suivre des mises en place et des enlèvements d'articles d'inventaire par des sujets dans une zone d'un espace réel. Une pluralité de caméras ayant des champs de vision se chevauchant produisent des séquences respectives d'images de champs de vision correspondants dans l'espace réel. Un système de traitement est couplé au système. Dans un mode de réalisation, le système de traitement comprend des moteurs de reconnaissance d'image recevant des séquences correspondantes d'images provenant de la pluralité des caméras. Les moteurs de reconnaissance d'image traitent les images dans les séquences correspondantes pour identifier des sujets représentés dans les images et générer des classifications des sujets identifiés. Le système traite les classifications de sujets identifiés pour des ensembles d'images dans les séquences d'images pour détecter des mises en place et des enlèvements d'articles d'inventaire sur des étagères par des sujets identifiés.
EP18844384.0A 2017-08-07 2018-07-26 Détection de mise en place et enlèvement d'article à l'aide d'une reconnaissance d'image Withdrawn EP3665648A4 (fr)

Applications Claiming Priority (6)

Application Number Priority Date Filing Date Title
US201762542077P 2017-08-07 2017-08-07
US15/847,796 US10055853B1 (en) 2017-08-07 2017-12-19 Subject identification and tracking using image recognition
US15/907,112 US10133933B1 (en) 2017-08-07 2018-02-27 Item put and take detection using image recognition
US15/945,466 US10127438B1 (en) 2017-08-07 2018-04-04 Predicting inventory events using semantic diffing
US15/945,473 US10474988B2 (en) 2017-08-07 2018-04-04 Predicting inventory events using foreground/background processing
PCT/US2018/043934 WO2019032305A2 (fr) 2017-08-07 2018-07-26 Détection de mise en place et enlèvement d'article à l'aide d'une reconnaissance d'image

Publications (2)

Publication Number Publication Date
EP3665648A2 true EP3665648A2 (fr) 2020-06-17
EP3665648A4 EP3665648A4 (fr) 2020-12-30

Family

ID=65271305

Family Applications (4)

Application Number Title Priority Date Filing Date
EP18843163.9A Withdrawn EP3665615A4 (fr) 2017-08-07 2018-07-26 Prédiction d'événements d'inventaire à l'aide d'une différenciation sémantique
EP18843486.4A Withdrawn EP3665647A4 (fr) 2017-08-07 2018-07-26 Prédiction d'événements d'inventaire au moyen d'un traitement d'avant-plan/arrière-plan
EP18844384.0A Withdrawn EP3665648A4 (fr) 2017-08-07 2018-07-26 Détection de mise en place et enlèvement d'article à l'aide d'une reconnaissance d'image
EP18844509.2A Withdrawn EP3665649A4 (fr) 2017-08-07 2018-07-26 Identification et suivi de sujet à l'aide d'une reconnaissance d'image

Family Applications Before (2)

Application Number Title Priority Date Filing Date
EP18843163.9A Withdrawn EP3665615A4 (fr) 2017-08-07 2018-07-26 Prédiction d'événements d'inventaire à l'aide d'une différenciation sémantique
EP18843486.4A Withdrawn EP3665647A4 (fr) 2017-08-07 2018-07-26 Prédiction d'événements d'inventaire au moyen d'un traitement d'avant-plan/arrière-plan

Family Applications After (1)

Application Number Title Priority Date Filing Date
EP18844509.2A Withdrawn EP3665649A4 (fr) 2017-08-07 2018-07-26 Identification et suivi de sujet à l'aide d'une reconnaissance d'image

Country Status (5)

Country Link
EP (4) EP3665615A4 (fr)
JP (4) JP7228569B2 (fr)
CA (4) CA3072058A1 (fr)
TW (1) TWI773797B (fr)
WO (4) WO2019032307A1 (fr)

Families Citing this family (103)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4410155A1 (fr) 2016-05-09 2024-08-07 Grabango Co. Système et procédé pour des applications commandées par vision par ordinateur dans un environnement
US10282621B2 (en) 2016-07-09 2019-05-07 Grabango Co. Remote state following device
WO2018148613A1 (fr) 2017-02-10 2018-08-16 Grabango Co. Expérience dynamique de caisse pour client dans un environnement d'achat automatisé
GB2560177A (en) 2017-03-01 2018-09-05 Thirdeye Labs Ltd Training a computational neural network
GB2560387B (en) 2017-03-10 2022-03-09 Standard Cognition Corp Action identification using neural networks
JP7165140B2 (ja) 2017-05-10 2022-11-02 グラバンゴ コーポレイション 効率的配置のための直列構成カメラアレイ
IL271528B2 (en) 2017-06-21 2024-08-01 Grabango Co Observed link of human activity in the video to a user account
US10650545B2 (en) 2017-08-07 2020-05-12 Standard Cognition, Corp. Systems and methods to check-in shoppers in a cashier-less store
US10474991B2 (en) 2017-08-07 2019-11-12 Standard Cognition, Corp. Deep learning-based store realograms
US11200692B2 (en) 2017-08-07 2021-12-14 Standard Cognition, Corp Systems and methods to check-in shoppers in a cashier-less store
US10474988B2 (en) 2017-08-07 2019-11-12 Standard Cognition, Corp. Predicting inventory events using foreground/background processing
US10853965B2 (en) 2017-08-07 2020-12-01 Standard Cognition, Corp Directional impression analysis using deep learning
US11023850B2 (en) 2017-08-07 2021-06-01 Standard Cognition, Corp. Realtime inventory location management using deep learning
US11232687B2 (en) 2017-08-07 2022-01-25 Standard Cognition, Corp Deep learning-based shopper statuses in a cashier-less store
US11250376B2 (en) 2017-08-07 2022-02-15 Standard Cognition, Corp Product correlation analysis using deep learning
US20190079591A1 (en) 2017-09-14 2019-03-14 Grabango Co. System and method for human gesture processing from video input
US10963704B2 (en) 2017-10-16 2021-03-30 Grabango Co. Multiple-factor verification for vision-based systems
US11481805B2 (en) 2018-01-03 2022-10-25 Grabango Co. Marketing and couponing in a retail environment using computer vision
US11257225B2 (en) 2019-10-25 2022-02-22 7-Eleven, Inc. Sensor mapping to a global coordinate system using homography
US11113837B2 (en) 2019-10-25 2021-09-07 7-Eleven, Inc. Sensor mapping to a global coordinate system
US11132550B2 (en) 2019-10-25 2021-09-28 7-Eleven, Inc. Detecting shelf interactions using a sensor array
US11080529B2 (en) 2019-10-25 2021-08-03 7-Eleven, Inc. Determining candidate object identities during image tracking
US11308630B2 (en) 2019-10-25 2022-04-19 7-Eleven, Inc. Auto-exclusion zone for contour-based object detection
US10769450B1 (en) 2019-10-25 2020-09-08 7-Eleven, Inc. Tracking positions using a scalable position tracking system
US11107226B2 (en) 2019-10-25 2021-08-31 7-Eleven, Inc. Object re-identification during image tracking
US11176686B2 (en) 2019-10-25 2021-11-16 7-Eleven, Inc. Image-based action detection using contour dilation
US10614318B1 (en) 2019-10-25 2020-04-07 7-Eleven, Inc. Sensor mapping to a global coordinate system using a marker grid
US11568554B2 (en) 2019-10-25 2023-01-31 7-Eleven, Inc. Contour-based detection of closely spaced objects
US11030756B2 (en) 2018-10-26 2021-06-08 7-Eleven, Inc. System and method for position tracking using edge computing
US11367124B2 (en) 2019-10-25 2022-06-21 7-Eleven, Inc. Detecting and identifying misplaced items using a sensor array
US11188763B2 (en) 2019-10-25 2021-11-30 7-Eleven, Inc. Topview object tracking using a sensor array
US10621444B1 (en) 2019-10-25 2020-04-14 7-Eleven, Inc. Action detection during image tracking
US11430046B2 (en) 2019-10-25 2022-08-30 7-Eleven, Inc. Identifying non-uniform weight objects using a sensor array
US11004219B1 (en) 2019-10-25 2021-05-11 7-Eleven, Inc. Vector-based object re-identification during image tracking
US10885642B1 (en) 2019-10-25 2021-01-05 7-Eleven, Inc. Scalable position tracking system for tracking position in large spaces
US10783762B1 (en) 2019-10-25 2020-09-22 7-Eleven, Inc. Custom rack for scalable position tracking system
US10943287B1 (en) 2019-10-25 2021-03-09 7-Eleven, Inc. Topview item tracking using a sensor array
US11062147B2 (en) 2019-10-25 2021-07-13 7-Eleven, Inc. Object assignment during image tracking
US10956777B1 (en) 2019-10-25 2021-03-23 7-Eleven, Inc. Shelf position calibration in a global coordinate system using a sensor array
US10789720B1 (en) 2019-10-25 2020-09-29 7-Eleven, Inc. Multi-camera image tracking on a global plane
US10878585B1 (en) 2019-10-25 2020-12-29 7-Eleven, Inc. Sensor array for scalable position tracking system
US11288518B2 (en) 2019-10-25 2022-03-29 7-Eleven, Inc. Tracking positions using a scalable position tracking system
US11288648B2 (en) 2018-10-29 2022-03-29 Grabango Co. Commerce automation for a fueling station
WO2020180815A1 (fr) 2019-03-01 2020-09-10 Grabango Co. Interface de caissier pour relier des clients à des données virtuelles
US11232575B2 (en) 2019-04-18 2022-01-25 Standard Cognition, Corp Systems and methods for deep learning-based subject persistence
DE102019209463A1 (de) * 2019-06-27 2020-12-31 Robert Bosch Gmbh Verfahren zur Bestimmung eines Vertrauenswertes eines Objektes einer Klasse
CN110503092B (zh) * 2019-07-22 2023-07-14 天津科技大学 基于领域适应的改进ssd监控视频目标检测方法
CN110705274B (zh) * 2019-09-06 2023-03-24 电子科技大学 基于实时学习的融合型词义嵌入方法
JP6982259B2 (ja) * 2019-09-19 2021-12-17 キヤノンマーケティングジャパン株式会社 情報処理装置、情報処理方法、プログラム
CN112560834B (zh) * 2019-09-26 2024-05-10 武汉金山办公软件有限公司 一种坐标预测模型生成方法、装置及图形识别方法、装置
US11893759B2 (en) 2019-10-24 2024-02-06 7-Eleven, Inc. Homography error correction using a disparity mapping
US11023740B2 (en) 2019-10-25 2021-06-01 7-Eleven, Inc. System and method for providing machine-generated tickets to facilitate tracking
US11501454B2 (en) 2019-10-25 2022-11-15 7-Eleven, Inc. Mapping wireless weight sensor array for item detection and identification
US10607080B1 (en) 2019-10-25 2020-03-31 7-Eleven, Inc. Feedback and training for a machine learning algorithm configured to determine customer purchases during a shopping session at a physical store
US11341569B2 (en) 2019-10-25 2022-05-24 7-Eleven, Inc. System and method for populating a virtual shopping cart based on video of a customer's shopping session at a physical store
US10861085B1 (en) 2019-10-25 2020-12-08 7-Eleven, Inc. Apparatus, system and method for populating a virtual shopping cart based on video of a customers shopping session at a physical store
US11023741B1 (en) 2019-10-25 2021-06-01 7-Eleven, Inc. Draw wire encoder based homography
US11893757B2 (en) 2019-10-25 2024-02-06 7-Eleven, Inc. Self-serve beverage detection and assignment
MX2022004898A (es) 2019-10-25 2022-05-16 7 Eleven Inc Deteccion de accion durante el seguimiento de imagenes.
US11017229B2 (en) 2019-10-25 2021-05-25 7-Eleven, Inc. System and method for selectively verifying algorithmically populated shopping carts
US11113541B2 (en) 2019-10-25 2021-09-07 7-Eleven, Inc. Detection of object removal and replacement from a shelf
US11450011B2 (en) 2019-10-25 2022-09-20 7-Eleven, Inc. Adaptive item counting algorithm for weight sensor using sensitivity analysis of the weight sensor
US11023728B1 (en) 2019-10-25 2021-06-01 7-Eleven, Inc. Machine learning algorithm trained to identify algorithmically populated shopping carts as candidates for verification
US11380091B2 (en) 2019-10-25 2022-07-05 7-Eleven, Inc. System and method for populating a virtual shopping cart based on a verification of algorithmic determinations of items selected during a shopping session in a physical store
US12062191B2 (en) 2019-10-25 2024-08-13 7-Eleven, Inc. Food detection using a sensor array
US11551454B2 (en) 2019-10-25 2023-01-10 7-Eleven, Inc. Homography error correction using marker locations
US11003918B1 (en) 2019-10-25 2021-05-11 7-Eleven, Inc. Event trigger based on region-of-interest near hand-shelf interaction
US10922555B1 (en) 2019-10-25 2021-02-16 7-Eleven, Inc. Customer-based video feed
US11386647B2 (en) 2019-10-25 2022-07-12 7-Eleven, Inc. System and method for processing a refund request arising from a shopping session in a cashierless store
US11587243B2 (en) 2019-10-25 2023-02-21 7-Eleven, Inc. System and method for position tracking using edge computing
US11887337B2 (en) 2019-10-25 2024-01-30 7-Eleven, Inc. Reconfigurable sensor array
US11100717B2 (en) 2019-10-25 2021-08-24 7-Eleven, Inc. System and method for presenting a virtual store shelf that emulates a physical store shelf
US11403852B2 (en) 2019-10-25 2022-08-02 7-Eleven, Inc. Object detection based on wrist-area region-of-interest
US11887372B2 (en) 2019-10-25 2024-01-30 7-Eleven, Inc. Image-based self-serve beverage detection and assignment
US11674792B2 (en) 2019-10-25 2023-06-13 7-Eleven, Inc. Sensor array with adjustable camera positions
US11557124B2 (en) 2019-10-25 2023-01-17 7-Eleven, Inc. Homography error correction
US11798065B2 (en) 2019-10-25 2023-10-24 7-Eleven, Inc. Tool for generating a virtual store that emulates a physical store
TWI703837B (zh) * 2019-11-29 2020-09-01 中華電信股份有限公司 用於管理雲端資料中心的管理系統和管理方法
TWI769420B (zh) * 2019-12-11 2022-07-01 財團法人工業技術研究院 智慧貨架圖產生方法與系統
JP7370845B2 (ja) * 2019-12-17 2023-10-30 東芝テック株式会社 販売管理装置及びその制御プログラム
US11093736B1 (en) * 2020-01-24 2021-08-17 Synchrony Bank Systems and methods for machine vision based object recognition
US11651516B2 (en) 2020-02-20 2023-05-16 Sony Group Corporation Multiple view triangulation with improved robustness to observation errors
JP7400532B2 (ja) * 2020-02-26 2023-12-19 株式会社村田製作所 複合炭素材およびその製造方法、リチウムイオン二次電池用負極活物質並びにリチウムイオン二次電池
JP7400533B2 (ja) * 2020-02-26 2023-12-19 株式会社村田製作所 複合炭素材およびその製造方法、電極作製用スラリー、電極塗膜並びにリチウムイオン二次電池
US11475595B2 (en) 2020-03-19 2022-10-18 Sony Corporation Extrinsic calibration of multi-camera system
CN111461141B (zh) * 2020-03-30 2023-08-29 歌尔科技有限公司 一种设备位姿计算方法装置及设备
JPWO2021230123A1 (fr) * 2020-05-12 2021-11-18
US11303853B2 (en) 2020-06-26 2022-04-12 Standard Cognition, Corp. Systems and methods for automated design of camera placement and cameras arrangements for autonomous checkout
US11361468B2 (en) 2020-06-26 2022-06-14 Standard Cognition, Corp. Systems and methods for automated recalibration of sensors for autonomous checkout
CN114120083A (zh) * 2020-08-12 2022-03-01 东芝泰格有限公司 图像识别装置及存储介质、图像识别方法
CN112508962B (zh) * 2020-09-27 2024-10-18 绍兴文理学院 基于时间关联图像序列的目标图像区域子序列分离方法
CN112767252B (zh) * 2021-01-26 2022-08-02 电子科技大学 一种基于卷积神经网络的图像超分辨率重建方法
EP4309120A4 (fr) * 2021-03-17 2024-07-24 Vispera Bilgi Teknolojileri Sanayi Ic Ve Dis Ticar A S Système et procédé de surveillance d'unités dans une armoire
CN117916742A (zh) * 2021-06-17 2024-04-19 Abb瑞士股份有限公司 用于基于神经网络输出来更新神经网络的训练的机器人系统和方法
JP2023031618A (ja) * 2021-08-25 2023-03-09 株式会社東芝 物品管理装置、物品管理方法、物品管理システム及びプログラム
FR3128048B1 (fr) * 2021-10-13 2024-09-20 Mo Ka Borne d’encaissement automatique intelligente
JP2023064427A (ja) * 2021-10-26 2023-05-11 富士通株式会社 推論プログラム、学習プログラム、推論方法および学習方法
CN114444895A (zh) * 2021-12-31 2022-05-06 深圳云天励飞技术股份有限公司 清洁质量评估方法及相关设备
JP7276535B1 (ja) * 2022-02-22 2023-05-18 富士通株式会社 情報処理プログラム、情報処理方法、および情報処理装置
CN114842045B (zh) * 2022-04-01 2024-04-16 深圳市九天睿芯科技有限公司 目标跟踪方法及装置
WO2024054784A1 (fr) * 2022-09-08 2024-03-14 Digimarc Corporation Procédés et agencements d'analyse d'images
CN116403269B (zh) * 2023-05-17 2024-03-26 智慧眼科技股份有限公司 一种遮挡人脸解析方法、系统、设备及计算机存储介质
CN118397568B (zh) * 2024-06-26 2024-09-03 汉朔科技股份有限公司 购物车的非正常购物行为检测方法、装置、系统及购物车

Family Cites Families (46)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3584334B2 (ja) * 1997-12-05 2004-11-04 オムロン株式会社 人間検出追跡システムおよび人間検出追跡方法
US7058204B2 (en) 2000-10-03 2006-06-06 Gesturetek, Inc. Multiple camera control system
US6678413B1 (en) * 2000-11-24 2004-01-13 Yiqing Liang System and method for object identification and behavior characterization using video analysis
US7688349B2 (en) * 2001-12-07 2010-03-30 International Business Machines Corporation Method of detecting and tracking groups of people
JP2004171241A (ja) 2002-11-20 2004-06-17 Casio Comput Co Ltd 不正監視システムおよびプログラム
JP2004171240A (ja) 2002-11-20 2004-06-17 Casio Comput Co Ltd 不正監視システムおよびプログラム
JP4125634B2 (ja) * 2003-05-26 2008-07-30 Ｎｅｃソフト株式会社 顧客情報収集管理方法及びそのシステム
KR100519782B1 (ko) * 2004-03-04 2005-10-07 삼성전자주식회사 스테레오 카메라를 이용한 사람 검출 방법 및 장치
JP2006126926A (ja) 2004-10-26 2006-05-18 Canon Inc 顧客データ作成システム、プログラムおよび方法
US20090041297A1 (en) * 2005-05-31 2009-02-12 Objectvideo, Inc. Human detection and tracking for security applications
GB0617604D0 (en) * 2006-09-07 2006-10-18 Zeroshift Ltd Inventory control system
US8189926B2 (en) * 2006-12-30 2012-05-29 Videomining Corporation Method and system for automatically analyzing categories in a physical space based on the visual characterization of people
JP2008176504A (ja) 2007-01-17 2008-07-31 Toshiba Corp 物体検出装置及びその方法
US7961946B2 (en) * 2007-05-15 2011-06-14 Digisensory Technologies Pty Ltd Method and system for background estimation in localization and tracking of objects in a smart video camera
US20090217315A1 (en) * 2008-02-26 2009-08-27 Cognovision Solutions Inc. Method and system for audience measurement and targeting media
JP4970195B2 (ja) * 2007-08-23 2012-07-04 株式会社日立国際電気 人物追跡システム、人物追跡装置および人物追跡プログラム
US8098888B1 (en) * 2008-01-28 2012-01-17 Videomining Corporation Method and system for automatic analysis of the trip of people in a retail space using multiple cameras
DE102008007199A1 (de) * 2008-02-01 2009-08-06 Robert Bosch Gmbh Maskierungsmodul für ein Videoüberwachungssystem, Verfahren zur Maskierung von selektierten Objekten sowie Computerprogramm
JP4585580B2 (ja) 2008-04-24 2010-11-24 東芝テック株式会社 人物動線追跡システム
JP2010002997A (ja) * 2008-06-18 2010-01-07 Toshiba Tec Corp 人物行動分析装置及び人物行動分析プログラム
US8009863B1 (en) * 2008-06-30 2011-08-30 Videomining Corporation Method and system for analyzing shopping behavior using multiple sensor tracking
US8180107B2 (en) * 2009-02-13 2012-05-15 Sri International Active coordinated tracking for multi-camera systems
JP2011209966A (ja) * 2010-03-29 2011-10-20 Sony Corp 画像処理装置および方法、並びにプログラム
US20110320322A1 (en) * 2010-06-25 2011-12-29 Symbol Technologies, Inc. Inventory monitoring using complementary modes for item identification
JP2012123667A (ja) * 2010-12-09 2012-06-28 Panasonic Corp 姿勢推定装置および姿勢推定方法
US9213405B2 (en) * 2010-12-16 2015-12-15 Microsoft Technology Licensing, Llc Comprehension and intent-based content for augmented reality displays
US8448056B2 (en) * 2010-12-17 2013-05-21 Microsoft Corporation Validation analysis of human target
JP2012147083A (ja) 2011-01-07 2012-08-02 Sony Corp 画像処理装置、画像処理方法、およびプログラム
US8890357B2 (en) 2011-01-17 2014-11-18 Balboa Water Group, Inc. Bathing system transformer device with first and second low voltage output power connections
US9117113B2 (en) * 2011-05-13 2015-08-25 Liberovision Ag Silhouette-based pose estimation
US9177195B2 (en) * 2011-09-23 2015-11-03 Shoppertrak Rct Corporation System and method for detecting, tracking and counting human objects of interest using a counting system and a data capture device
US8630457B2 (en) * 2011-12-15 2014-01-14 Microsoft Corporation Problem states for pose tracking pipeline
US9338409B2 (en) * 2012-01-17 2016-05-10 Avigilon Fortress Corporation System and method for home health care monitoring
US9124778B1 (en) 2012-08-29 2015-09-01 Nomi Corporation Apparatuses and methods for disparity-based tracking and analysis of objects in a region of interest
JP2014154037A (ja) 2013-02-12 2014-08-25 Toshiba Tec Corp 画像読取装置
US10176456B2 (en) * 2013-06-26 2019-01-08 Amazon Technologies, Inc. Transitioning items from a materials handling facility
US10268983B2 (en) * 2013-06-26 2019-04-23 Amazon Technologies, Inc. Detecting item interaction and movement
JP5632512B1 (ja) 2013-07-02 2014-11-26 パナソニック株式会社 人物行動分析装置、人物行動分析システムおよび人物行動分析方法、ならびに監視装置
US10290031B2 (en) * 2013-07-24 2019-05-14 Gregorio Reid Method and system for automated retail checkout using context recognition
JP5834196B2 (ja) 2014-02-05 2015-12-16 パナソニックＩｐマネジメント株式会社 モニタリング装置、モニタリングシステムおよびモニタリング方法
US9262681B1 (en) * 2014-11-19 2016-02-16 Amazon Technologies, Inc. Probabilistic registration of interactions, actions or activities from multiple views
JP2016144049A (ja) * 2015-02-02 2016-08-08 キヤノン株式会社 画像処理装置、画像処理方法、およびプログラム
JP5988225B2 (ja) 2015-02-25 2016-09-07 パナソニックＩｐマネジメント株式会社 モニタリング装置およびモニタリング方法
GB2542118B (en) * 2015-09-04 2021-05-19 Toshiba Europe Ltd A method, apparatus, system, and computer readable medium for detecting change to a structure
JP6590609B2 (ja) 2015-09-15 2019-10-16 キヤノン株式会社 画像解析装置及び画像解析方法
WO2017123920A1 (fr) * 2016-01-14 2017-07-20 RetailNext, Inc. Détection, suivi et comptage d'objets dans des vidéos

Also Published As

Publication number Publication date
EP3665647A4 (fr) 2021-01-06
EP3665649A1 (fr) 2020-06-17
TW201911119A (zh) 2019-03-16
JP7181922B2 (ja) 2022-12-01
WO2019032307A1 (fr) 2019-02-14
TWI773797B (zh) 2022-08-11
CA3072063A1 (fr) 2019-02-14
JP2020530170A (ja) 2020-10-15
CA3072058A1 (fr) 2019-02-14
WO2019032305A2 (fr) 2019-02-14
JP2020530168A (ja) 2020-10-15
WO2019032306A9 (fr) 2020-03-19
EP3665615A4 (fr) 2020-12-30
CA3072056A1 (fr) 2019-02-14
WO2019032304A1 (fr) 2019-02-14
JP2020530167A (ja) 2020-10-15
JP7191088B2 (ja) 2022-12-16
WO2019032305A3 (fr) 2019-03-21
WO2019032306A1 (fr) 2019-02-14
JP2021503636A (ja) 2021-02-12
JP7228569B2 (ja) 2023-02-24
CA3072062A1 (fr) 2019-02-14
EP3665648A4 (fr) 2020-12-30
EP3665649A4 (fr) 2021-01-06
EP3665615A1 (fr) 2020-06-17
JP7208974B2 (ja) 2023-01-19
EP3665647A1 (fr) 2020-06-17

Similar Documents

Publication Publication Date Title
US12026665B2 (en) Identifying inventory items using multiple confidence levels
US10127438B1 (en) Predicting inventory events using semantic diffing
US10133933B1 (en) Item put and take detection using image recognition
JP7228569B2 (ja) 画像認識を用いた被写体識別及び追跡
US10055853B1 (en) Subject identification and tracking using image recognition
US20220147913A1 (en) Inventory tracking system and method that identifies gestures of subjects holding inventory items
EP3844704A1 (fr) Reçus numériques actionnables fondés sur l'apprentissage profond pour vente sans caisse
WO2020023799A1 (fr) Analayse de corrélation de produits à l'aide d'un apprentissage profond
EP4172939A1 (fr) Systèmes et procédés de réétalonnage automatisé de capteurs pour passages en caisse en autonomie
US20210350555A1 (en) Systems and methods for detecting proximity events

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20200222

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

AX Request for extension of the european patent

Extension state: BA ME

RIN1 Information on inventor provided before grant (corrected)

Inventor name: NOVAK, JOHN F.

Inventor name: DORMAN, KYLE E.

Inventor name: FISHER, JORDAN E.

Inventor name: FISCHETTI, DANIEL L.

Inventor name: OGLE, BRANDON L.

Inventor name: LASHERAS, JUAN C.

Inventor name: KIHARA, KENNETH S.

RIN1 Information on inventor provided before grant (corrected)

Inventor name: DORMAN, KYLE E.

Inventor name: LASHERAS, JUAN C.

Inventor name: FISHER, JORDAN E.

Inventor name: KIHARA, KENNETH S.

Inventor name: OGLE, BRANDON L.

Inventor name: FISCHETTI, DANIEL L.

Inventor name: NOVAK, JOHN F.

RIN1 Information on inventor provided before grant (corrected)

Inventor name: LASHERAS, JUAN C.

Inventor name: FISCHETTI, DANIEL L.

Inventor name: KIHARA, KENNETH S.

Inventor name: FISHER, JORDAN E.

Inventor name: DORMAN, KYLE E.

Inventor name: NOVAK, JOHN F.

Inventor name: OGLE, BRANDON L.

DAV Request for validation of the european patent (deleted)
DAX Request for extension of the european patent (deleted)
REG Reference to a national code

Ref country code: DE

Ref legal event code: R079

Free format text: PREVIOUS MAIN CLASS: G06T0007292000

Ipc: G06K0009000000

A4 Supplementary search report drawn up and despatched

Effective date: 20201202

RIC1 Information provided on ipc code assigned before grant

Ipc: G06Q 20/20 20120101ALI20201126BHEP

Ipc: G06K 9/00 20060101AFI20201126BHEP

Ipc: G06K 9/40 20060101ALI20201126BHEP

Ipc: G06N 3/08 20060101ALI20201126BHEP

Ipc: G06K 9/62 20060101ALI20201126BHEP

Ipc: G06Q 30/06 20120101ALI20201126BHEP

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

17Q First examination report despatched

Effective date: 20220421

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20231219