CN109657537A - Image-recognizing method, system and electronic equipment based on target detection - Google Patents
Image-recognizing method, system and electronic equipment based on target detection Download PDFInfo
- Publication number
- CN109657537A CN109657537A CN201811309239.1A CN201811309239A CN109657537A CN 109657537 A CN109657537 A CN 109657537A CN 201811309239 A CN201811309239 A CN 201811309239A CN 109657537 A CN109657537 A CN 109657537A
- Authority
- CN
- China
- Prior art keywords
- image
- gesture
- recognizing method
- target
- key point
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Abstract
The application is about a kind of image-recognizing method based on target detection characterized by comprising obtains image to be processed;According to the target convolutional neural networks for constructing and being trained acquisition in advance, the gesture in image is identified;Obtain the location information of the classification of the gesture and the key point of the gesture.The image-recognizing method has carried out deeper parsing to image, to realize more convenient application image information abundant, realizes the human-computer interaction mode of more flexible multiplicity.
Description
Technical field
This application involves technical field of image processing more particularly to a kind of image-recognizing method based on target detection and it is
System.
Background technique
In recent years, deep learning has obtained extensively in related fieldss such as video image processing, speech recognition, natural language processings
In terms of general application, especially image procossing, the technology of image recognition is quickly grown.In image procossing, target detection (object
Detection) refer to specific object target in concern picture, when inputting a picture, which identifies and export this
The location information (bounding box) and classification information of target in picture.
Existing object detection method is broadly divided into two classes, and one kind is two stage target detection model, is also based on area
The recognition methods in domain, including R-CNN, Fast R-CNN, Faster R-CNN etc..Detection process is divided into two stages, first needle
Multiple alternative frames (region proposals) that may include target are extracted to input figure, then calculate multiple alternative frames again
Convolutional neural networks (CNN) feature, classifies to each alternative frame.In addition a kind of method, is the target detection mould of single phase
Type, this method are intended to directly acquire prediction result to an input picture, without being extracted based on the alternative frame in region method
Process, this method is also referred to as without area recognizing method, including SSD, yolo method etc..SSD full name Single Shot
Multibox Detector.Single shot refers to single phase object detection method, and multibox detectior refers to can
With the detection of more frames.For a picture, the detection block of the exportable target of SSD and the classification of target.
Existing SSD scheme directly returns out the position of picture target by one picture of input, and to the target into
Row classification.Such as the application scenarios in detection gesture target, SSD can only detect the position frame of gesture and the classification of gesture
(such as victory), but the position (such as position of gesture index finger tip) of gesture key point can not be returned out.It can not obtain
Gesture key point (for example, index finger tip) is taken, to carry out deeper parsing and application according to key point, is realized cleverer
The human-computer interaction mode of multiplicity living.
Summary of the invention
Present invention finds the Limited information that existing image recognition obtains, the to a certain degree upper limit in the course of the research
Made its application range, practical value is low, can not more convenient deeper realization human-computer interaction, therefore, to overcome related skill
The problem of art, the application disclose a kind of image-recognizing method based on target detection, system, electronic equipment and storage and are situated between
Matter.
According to the embodiment of the present application in a first aspect, providing a kind of image-recognizing method based on target detection, feature
It is, comprising:
Obtain image to be processed;
According to the target convolutional neural networks for constructing and being trained acquisition in advance, the hand in image is identified;
Obtain the location information of the classification of the gesture and the key point of the gesture.
Preferably, the target convolutional neural networks are obtained by following steps:
Obtain training image;
The training image is labeled;
Construction exports the initial convolutional neural networks of the label information of image for input information;
Using training image as input, the mark of combined training image is trained optimization to initial convolution mind grade network,
Obtain the target convolutional neural networks.
Preferably, the location information marked including the classification of gesture and the key point of gesture.
Preferably, the mark further includes the number of gesture.
Preferably, the key point of the gesture include finger fingertip, in the centre of the palm to a little less.
Preferably, coordinate representation of the location information of the key point by the key point in the picture.
It preferably, include multiple gestures in described image.
Preferably, the target convolutional neural networks are detected using Analysis On Multi-scale Features.
Preferably, described image recognition methods includes single phase object detection method.
According to the second aspect of the embodiment of the present application, a kind of image identification system is provided characterized by comprising
Module is obtained, for obtaining image to be processed;
Processing module, for the image to be processed to be input to the target convolution for constructing and being trained in advance acquisition
Neural network identifies the gesture in image, and obtains the location information of the classification of the gesture and the key point of the gesture;
Output module, the location information of the key point for exporting gesture classification and the gesture in described image.
Preferably, the processing module handles multiple continuous images in time, to obtain the fortune of the key point
Dynamic path.
Preferably, described image identification device is according to image recognition as a result, increasing special efficacy for described image.
According to the third aspect of the embodiment of the present application, a kind of electronic equipment is provided characterized by comprising
One or more processors;
Memory;
One or more application program, wherein one or more of application programs are stored in the memory simultaneously
It is configured as being executed by one or more of processors, one or more of programs are configured to carry out above-mentioned any one
The image-recognizing method.
According to the fourth aspect of the embodiment of the present application, a kind of non-transitorycomputer readable storage medium is provided, when described
When instruction in storage medium is executed by the processor of electronic equipment, so that electronic equipment is able to carry out a kind of image recognition side
Method, the method includes image-recognizing methods described in above-mentioned any one.
The technical solution that embodiments herein provides can include the following benefits:
1) present applicant proposes a kind of image-recognizing methods based on target detection, by increasing and marking in the training stage
More useful informations, to predict and obtain the key point location information of gesture in the image intentionally got, so as to utilize
The information expands more application scenarios, and further, the position that according to the location information of the key point, can also track target is moved
It is dynamic, the motion profile etc. of key point is obtained, when to provide more favorable information for various applications, such as to record small video,
This programme can not only obtain hand gesture location and classification, can also be tracked at the same time by returning index finger key point position
Firefinger movement situation, to targetedly add some special efficacys.
2) image-recognizing method of the application is easily achieved, and uses the target detection model of single phase, simplifies knowledge
Other step, and detected by Analysis On Multi-scale Features figure, large-scale characteristics figure can divide more junior units, each unit
Priori frame is smaller, for detecting Small object.The characteristic pattern of small scale can divide bigger unit, each unit priori frame
Scale is bigger, for detecting big target.The position of target in picture can be directly obtained, is determined by inputting a picture
Destination number simultaneously classifies to the target.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The application can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the application
Example, and together with specification it is used to explain the principle of the application.
Fig. 1 is the flow chart of the image-recognizing method shown according to an exemplary embodiment based on target detection;
Fig. 2 is the step schematic diagram shown according to an exemplary embodiment for obtaining target convolutional neural networks;
Fig. 3 is the schematic diagram of the location information of gesture key point in acquisition image shown according to an exemplary embodiment;
Fig. 4 is the schematic network structure of initial convolutional neural networks shown according to an exemplary embodiment;
Fig. 5 is the image identification system schematic diagram shown according to an exemplary embodiment based on target detection;
Fig. 6 is the schematic diagram of image recognition apparatus shown according to an exemplary embodiment.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
Fig. 1 is the flow chart of the image-recognizing method based on target detection shown accoding to exemplary embodiment, specific to wrap
Include following steps:
In step s101, image to be processed is obtained;
In step s 102, it according to the target convolutional neural networks for constructing and being trained acquisition in advance, identifies to be processed
Image in gesture;
In step s 103, the location information of the classification of the gesture and the key point of the gesture is obtained.
In one embodiment of the invention, firstly, obtaining image to be processed;Then, according to constructing and carry out in advance
The target convolutional neural networks that training obtains identify in image whether include gesture;Finally obtain the number of contained gesture in image
Classification belonging to amount and each gesture, and the key point (for example, finger tip of index finger in gesture) of each gesture is obtained in described image
In location information etc..Target convolutional neural networks of the invention are based on a kind of object detection method, which uses
The target detection model of single phase, and image is detected by Analysis On Multi-scale Features figure.May be implemented in identification image in whether
It is other simultaneously comprising gesture and gesture class, obtain location information of the key point of gesture in described image.Of course, as carried out
Gesture is not included in the image of identification, it is concluded that the information without gesture.Specifically, the target detection side SSD is used in the present invention
Method, SSD full name are Single Shot Multibox Detector, refer to single phase more frame detections.
Fig. 2 is the step schematic diagram shown according to an exemplary embodiment for obtaining target convolutional neural networks, specific to wrap
Include following steps:
In step s 201, training image is obtained, further, data cleansing can be carried out to the training image of acquisition, with
Delete the duplicate message in image;Image after data cleansing is, for example, 1000, and image size is, for example, 640*480.
In step S202, the training image after data cleansing is labeled, constructs training set;The mark packet
The quantity of gesture is included, the position of the position of the key point of the type and gesture of gesture, the key point passes through the key point
Coordinate in described image is indicated.
In step S203, construction exports the initial convolutional neural networks of the label information of gesture, institute for input information
The label information for stating gesture includes n multi-C vector, and n is the quantity of gesture in described image.The multi-C vector is, for example, 6 dimensions
Vector can recognize 4 kinds of gesture classifications, and 2 dimensions are the coordinates of the key point of gesture in 6 dimensional vectors, indicate that the key point of gesture is being schemed
Coordinate value as in;4 dimensions are gesture classifications, and belonging to the category is 1, and being not belonging to the category is 0.The initial volume in the present invention
It is, for example, SSD object detection method that product neural network, which is based on target detection image-recognizing method,.The network structure for example comprising
Multiple concatenated convolutional layers and full articulamentum.
In step S204, the training sample (training image after mark) in training set is used as and is inputted, described in input
Initial convolutional neural networks are trained, and update the weight of the initial convolutional neural networks in the training process, are obtained described
Target convolutional neural networks.
Fig. 3 is the schematic diagram of the location information of gesture key point in acquisition image shown according to an exemplary embodiment,
Include 100 and 200 total 2 gestures in figure, is respectively x-axis and y-axis with two sides orthogonal in image, establishes right angle
Coordinate system includes its picture number in described image title, and for example, 1.jpg, corresponding mark file is 1.txt.It chooses
The position of key point (index finger tip) coordinate representation key point of gesture respectively indicates gesture using the coordinate of A point and B point
100, the key point position of gesture 200, the then content of 1.txt are as follows:
1 XA,YA,Z1,Z2,Z3,Z4
1 XB,YB,Z1,Z2,Z3,Z4
Wherein, first be classified as picture number, behind two coordinates for being classified as key point, the line number of data indicates institute in picture
Including gesture quantity, Z1, Z2, Z3, Z4 respectively indicate the classification that the gesture adheres to separately.
As shown in figure 4, the initial convolutional neural networks in the present invention are based on a kind of SSD object detection method, the net
Network structure is for example divided into two parts, and front is standard network (eliminating the relevant layer of classification) for image classification, behind
Network be Analysis On Multi-scale Features mapping layer for detection, to reach the different size of target of detection.Gle shot refers to list
Phase targets detection method, multibox detectior refers to can more frame detections.For a picture, SSD exports target
Detection block and target classification.And SSD is detected using Analysis On Multi-scale Features figure, large-scale characteristics figure can divide more
The priori frame of more junior units, each unit is smaller, for detecting Small object.The characteristic pattern of small scale can divide bigger
The scale of unit, each unit priori frame is bigger, for detecting big target.Further, SSD directlys adopt convolution to difference
Characteristic pattern detected.The different priori frame of length-width ratio can be arranged in SSD on each unit, predict boundingbox when
Time can be based on the priori frame of these units.Also the priori frame of matching realistic objective body form can be found when training.The SSD
Scheme directly returns out the position of picture target, that is, bounding box, and to the target by one picture of input
Classify.For example, consider detection gesture target application scenarios, conventional SSD can only detect gesture position frame and
The classification (such as victory) of gesture, but the position of gesture key point can not be returned out.But method of the invention is exactly base
In SSD single phase object detection method, not only gesture is predicted and classified to the position frame of gesture, can also obtain gesture food
Refer to the position of finger tip (key point).
Fig. 5 is the image identification system schematic diagram shown according to an exemplary embodiment based on target detection, such as Fig. 5 institute
Show, which includes: image collection module 301, image processing module 302, output module 303 and special efficacy mould
Block 304.
Image collection module 301: for carrying out image acquisition, for example, camera, for obtaining image to be processed.
Image processing module 302: it for being identified according to be processed image of the target convolutional neural networks to acquisition, obtains
Take the label information of gesture in image.
Output module 303: according to the label information of gesture in described image, data output is carried out, the data of output include:
The quantity of gesture in image, the classification of each gesture, the key point position of each gesture.Further, the output module 303 may be used also
Realize the dynamic following output of key point position, such as image to be processed is the image continuously acquired in certain period of time, described defeated
Module can also export the motion profile of each key point position according to the key point position of gesture each in image out.
Interactive module 304: according to output module 303 export as a result, key point position in the picture is acted, example
Special efficacy for example is added to the key point position in image, of course, the special efficacy can also follow the mobile progress of the key point
It is mobile, realize the tracking effect of addition special efficacy.Further, the FX Module 304 can also be defeated according to the output module 303
Out as a result, obtain the other variation of gesture class in image, and replace special efficacy.
In one embodiment of the invention, target convolutional neural networks are based on the initial convolution neural network and cross instruction
White silk optimizes, and using the training sample in training set as input, is trained to the initial convolutional neural networks, in training
The weight of the initial convolutional neural networks is updated in the process, to obtain the target convolutional neural networks.
Fig. 6 is the schematic diagram of image recognition apparatus shown according to an exemplary embodiment.Equipment shown in Fig. 6 is only
One example, should not function to the embodiment of the present invention and use scope constitute any restrictions.
With reference to Fig. 6, which includes the processor 401, memory 402 and input connected by bus
Output device 403.Memory 402 includes read-only memory (ROM) and random access storage device (RAM), storage in memory 402
There are various computer instructions and data needed for executing system function, processor 401 reads various computers from memory 402
Instruction is to execute various movements appropriate and processing.Input/output unit includes the importation of keyboard, mouse etc.;Including such as
The output par, c of cathode-ray tube (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section including hard disk etc.;
And the communications portion of the network interface card including LAN card, modem etc..
Memory 402 is also stored with computer instruction below and is advised with the image-recognizing method for completing the embodiment of the present invention
Fixed operation: image to be processed is obtained;According to the target convolutional neural networks for constructing and being trained acquisition in advance, identification figure
Hand as in;Obtain the location information of the classification of the gesture and the key point of the gesture.
Correspondingly, the embodiment of the present invention provides a kind of computer readable storage medium, which deposits
Computer instruction is contained, the computer instruction is performed the operation for realizing above-mentioned image-recognizing method defined.
The application also provides computer program product, including computer program product, and the computer program includes program
Instruction, when described program instruction is executed by electronic equipment, the step of making the electronic equipment execute above-mentioned instant communicating method.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the application
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the application, these modifications, purposes or
Person's adaptive change follows the general principle of the application and including the undocumented common knowledge in the art of the application
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the application are by following
Claim is pointed out.
It should be understood that the application is not limited to the precise structure that has been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.Scope of the present application is only limited by the accompanying claims.
Claims (14)
1. a kind of image-recognizing method based on target detection characterized by comprising
Obtain image to be processed;
According to the target convolutional neural networks for constructing and being trained acquisition in advance, the gesture in image is identified;
Obtain the location information of the classification of the gesture and the key point of the gesture.
2. the image-recognizing method according to claim 1 based on target detection, which is characterized in that obtained by following steps
The target convolutional neural networks:
Obtain training image;
The training image is labeled;
Construction exports the initial convolutional neural networks of the label information of image for input information;
Using training image as input, the mark of combined training image is trained optimization to initial convolution mind grade network, obtains
The target convolutional neural networks.
3. the image-recognizing method according to claim 2 based on target detection, which is characterized in that the mark includes hand
The location information of the key point of the classification and gesture of gesture.
4. the image-recognizing method according to claim 3 based on target detection, which is characterized in that the mark further includes
The number of gesture.
5. the image-recognizing method according to claim 1 based on target detection, which is characterized in that the key of the gesture
Point is including extremely a little less in finger fingertip, the centre of the palm.
6. the image-recognizing method according to claim 1 based on target detection, which is characterized in that the position of the key point
Confidence ceases the coordinate representation by the key point in the picture.
7. the image-recognizing method according to claim 1 based on target detection, which is characterized in that include in described image
Multiple gestures.
8. the image-recognizing method according to claim 1 based on target detection, which is characterized in that the target convolution mind
It is detected through network using Analysis On Multi-scale Features.
9. the image-recognizing method according to claim 1 based on target detection, which is characterized in that described image identification side
Method includes single phase object detection method.
10. a kind of image identification system characterized by comprising
Image collection module, for obtaining image to be processed;
Image processing module, for the image to be processed to be input to the target convolution for constructing and being trained in advance acquisition
Neural network identifies the gesture in image, and obtains the location information of the classification of the gesture and the key point of the gesture;
Output module, the location information of the key point for exporting gesture classification and the gesture in described image.
11. image identification system according to claim 10, which is characterized in that the processing module processing connects in time
Multiple continuous images, to obtain the motion path of the key point.
12. image identification system according to claim 10, which is characterized in that it further include FX Module, the special efficacy mould
Root tuber is according to image recognition as a result, increasing special efficacy for described image.
13. a kind of electronic equipment characterized by comprising
One or more processors;
Memory;
One or more application program, wherein one or more of application programs are stored in the memory and are matched
It is set to and is executed by one or more of processors, it is any that one or more of programs are configured to carry out claim 1-9
Image-recognizing method described in one.
14. a kind of non-transitorycomputer readable storage medium, when the instruction in the storage medium is by the processing of electronic equipment
When device executes, so that electronic equipment is able to carry out a kind of image-recognizing method, the method includes the claims 1-9 is any
Image-recognizing method described in one.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811309239.1A CN109657537A (en) | 2018-11-05 | 2018-11-05 | Image-recognizing method, system and electronic equipment based on target detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811309239.1A CN109657537A (en) | 2018-11-05 | 2018-11-05 | Image-recognizing method, system and electronic equipment based on target detection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109657537A true CN109657537A (en) | 2019-04-19 |
Family
ID=66110216
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811309239.1A Pending CN109657537A (en) | 2018-11-05 | 2018-11-05 | Image-recognizing method, system and electronic equipment based on target detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109657537A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110084204A (en) * | 2019-04-29 | 2019-08-02 | 北京字节跳动网络技术有限公司 | Image processing method, device and electronic equipment based on target object posture |
CN110287891A (en) * | 2019-06-26 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Gestural control method, device and electronic equipment based on human body key point |
CN110334736A (en) * | 2019-06-03 | 2019-10-15 | 北京大米科技有限公司 | Image-recognizing method, device, electronic equipment and medium |
CN110807410A (en) * | 2019-10-30 | 2020-02-18 | 北京百度网讯科技有限公司 | Key point positioning method and device, electronic equipment and storage medium |
CN111126460A (en) * | 2019-12-10 | 2020-05-08 | 福建省高速公路集团有限公司 | Pavement disease automatic inspection method, medium, equipment and device based on artificial intelligence |
CN111625157A (en) * | 2020-05-20 | 2020-09-04 | 北京百度网讯科技有限公司 | Fingertip key point detection method, device, equipment and readable storage medium |
CN111985184A (en) * | 2020-06-30 | 2020-11-24 | 上海翎腾智能科技有限公司 | Auxiliary writing font copying method, system and device based on AI vision |
CN112418200A (en) * | 2021-01-25 | 2021-02-26 | 成都点泽智能科技有限公司 | Object detection method and device based on thermal imaging and server |
WO2021098587A1 (en) * | 2019-11-20 | 2021-05-27 | Oppo广东移动通信有限公司 | Gesture analysis method, apparatus and device, and computer-readable storage medium |
CN112949437A (en) * | 2021-02-21 | 2021-06-11 | 深圳市优必选科技股份有限公司 | Gesture recognition method, gesture recognition device and intelligent equipment |
CN114144105A (en) * | 2019-05-07 | 2022-03-04 | 伊利诺斯工具制品有限公司 | System for optical laundry recognition in a dishwasher, method for optical laundry recognition, dishwasher with such a system and method for operating such a dishwasher |
EP4181015A4 (en) * | 2020-07-29 | 2023-11-22 | Huawei Technologies Co., Ltd. | Method and device for recognizing handheld object |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104656878A (en) * | 2013-11-19 | 2015-05-27 | 华为技术有限公司 | Method, device and system for recognizing gesture |
CN107168527A (en) * | 2017-04-25 | 2017-09-15 | 华南理工大学 | The first visual angle gesture identification and exchange method based on region convolutional neural networks |
CN108062526A (en) * | 2017-12-15 | 2018-05-22 | 厦门美图之家科技有限公司 | A kind of estimation method of human posture and mobile terminal |
CN108133220A (en) * | 2016-11-30 | 2018-06-08 | 北京市商汤科技开发有限公司 | Model training, crucial point location and image processing method, system and electronic equipment |
CN108229318A (en) * | 2017-11-28 | 2018-06-29 | 北京市商汤科技开发有限公司 | The training method and device of gesture identification and gesture identification network, equipment, medium |
CN108229509A (en) * | 2016-12-16 | 2018-06-29 | 北京市商汤科技开发有限公司 | For identifying object type method for distinguishing and device, electronic equipment |
CN108509839A (en) * | 2018-02-02 | 2018-09-07 | 东华大学 | One kind being based on the efficient gestures detection recognition methods of region convolutional neural networks |
-
2018
- 2018-11-05 CN CN201811309239.1A patent/CN109657537A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104656878A (en) * | 2013-11-19 | 2015-05-27 | 华为技术有限公司 | Method, device and system for recognizing gesture |
CN108133220A (en) * | 2016-11-30 | 2018-06-08 | 北京市商汤科技开发有限公司 | Model training, crucial point location and image processing method, system and electronic equipment |
CN108229509A (en) * | 2016-12-16 | 2018-06-29 | 北京市商汤科技开发有限公司 | For identifying object type method for distinguishing and device, electronic equipment |
CN107168527A (en) * | 2017-04-25 | 2017-09-15 | 华南理工大学 | The first visual angle gesture identification and exchange method based on region convolutional neural networks |
CN108229318A (en) * | 2017-11-28 | 2018-06-29 | 北京市商汤科技开发有限公司 | The training method and device of gesture identification and gesture identification network, equipment, medium |
CN108062526A (en) * | 2017-12-15 | 2018-05-22 | 厦门美图之家科技有限公司 | A kind of estimation method of human posture and mobile terminal |
CN108509839A (en) * | 2018-02-02 | 2018-09-07 | 东华大学 | One kind being based on the efficient gestures detection recognition methods of region convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
郭川磊,何嘉: "基于转置卷积操作改进的单阶段多边框目标检测方法", 《计算机应用》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110084204A (en) * | 2019-04-29 | 2019-08-02 | 北京字节跳动网络技术有限公司 | Image processing method, device and electronic equipment based on target object posture |
CN110084204B (en) * | 2019-04-29 | 2020-11-24 | 北京字节跳动网络技术有限公司 | Image processing method and device based on target object posture and electronic equipment |
CN114144105A (en) * | 2019-05-07 | 2022-03-04 | 伊利诺斯工具制品有限公司 | System for optical laundry recognition in a dishwasher, method for optical laundry recognition, dishwasher with such a system and method for operating such a dishwasher |
CN110334736A (en) * | 2019-06-03 | 2019-10-15 | 北京大米科技有限公司 | Image-recognizing method, device, electronic equipment and medium |
CN110287891A (en) * | 2019-06-26 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Gestural control method, device and electronic equipment based on human body key point |
CN110807410A (en) * | 2019-10-30 | 2020-02-18 | 北京百度网讯科技有限公司 | Key point positioning method and device, electronic equipment and storage medium |
CN110807410B (en) * | 2019-10-30 | 2022-09-06 | 北京百度网讯科技有限公司 | Key point positioning method and device, electronic equipment and storage medium |
WO2021098587A1 (en) * | 2019-11-20 | 2021-05-27 | Oppo广东移动通信有限公司 | Gesture analysis method, apparatus and device, and computer-readable storage medium |
CN111126460A (en) * | 2019-12-10 | 2020-05-08 | 福建省高速公路集团有限公司 | Pavement disease automatic inspection method, medium, equipment and device based on artificial intelligence |
CN111126460B (en) * | 2019-12-10 | 2024-04-02 | 福建省高速公路集团有限公司 | Automatic pavement disease inspection method, medium, equipment and device based on artificial intelligence |
CN111625157B (en) * | 2020-05-20 | 2021-09-17 | 北京百度网讯科技有限公司 | Fingertip key point detection method, device, equipment and readable storage medium |
CN111625157A (en) * | 2020-05-20 | 2020-09-04 | 北京百度网讯科技有限公司 | Fingertip key point detection method, device, equipment and readable storage medium |
CN111985184A (en) * | 2020-06-30 | 2020-11-24 | 上海翎腾智能科技有限公司 | Auxiliary writing font copying method, system and device based on AI vision |
EP4181015A4 (en) * | 2020-07-29 | 2023-11-22 | Huawei Technologies Co., Ltd. | Method and device for recognizing handheld object |
CN112418200A (en) * | 2021-01-25 | 2021-02-26 | 成都点泽智能科技有限公司 | Object detection method and device based on thermal imaging and server |
CN112949437A (en) * | 2021-02-21 | 2021-06-11 | 深圳市优必选科技股份有限公司 | Gesture recognition method, gesture recognition device and intelligent equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109657537A (en) | Image-recognizing method, system and electronic equipment based on target detection | |
Kapidis et al. | Egocentric hand track and object-based human action recognition | |
US20200226786A1 (en) | Detecting pose using floating keypoint(s) | |
EP3005224A2 (en) | Gesture tracking and classification | |
KR20150108888A (en) | Part and state detection for gesture recognition | |
CN110462645A (en) | Sensor data processor with updating ability | |
Jiang et al. | Online robust action recognition based on a hierarchical model | |
US9734435B2 (en) | Recognition of hand poses by classification using discrete values | |
CN111553326B (en) | Hand motion recognition method and device, electronic equipment and storage medium | |
Al-Azzoa et al. | Human related-health actions detection using Android Camera based on TensorFlow Object Detection API | |
CN106537387B (en) | Retrieval/storage image associated with event | |
US20230020965A1 (en) | Method and apparatus for updating object recognition model | |
CN112052746A (en) | Target detection method and device, electronic equipment and readable storage medium | |
CN115699102A (en) | Tracking multiple objects in a video stream using occlusion aware single object tracking | |
CN111507285A (en) | Face attribute recognition method and device, computer equipment and storage medium | |
Liu et al. | Target recognition of sport athletes based on deep learning and convolutional neural network | |
CN115131604A (en) | Multi-label image classification method and device, electronic equipment and storage medium | |
Jiang et al. | independent hand gesture recognition with Kinect | |
Dwivedi et al. | Orientation invariant skeleton feature (oisf): a new feature for human activity recognition | |
Ansar et al. | Robust hand gesture tracking and recognition for healthcare via Recurent neural network | |
Gheitasi et al. | Estimation of hand skeletal postures by using deep convolutional neural networks | |
Ni et al. | Multiple granularity modeling: A coarse-to-fine framework for fine-grained action analysis | |
Yadav et al. | End-to-end bare-hand localization system for human–computer interaction: a comprehensive analysis and viable solution | |
Gavrilescu | Recognizing human gestures in videos by modeling the mutual context of body position and hands movement | |
CN115223239A (en) | Gesture recognition method and system, computer equipment and readable storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190419 |