CN113781435A - Cigarette packet appearance defect detection method based on YOLOV5 network - Google Patents

Cigarette packet appearance defect detection method based on YOLOV5 network Download PDF

Info

Publication number
CN113781435A
CN113781435A CN202111062913.2A CN202111062913A CN113781435A CN 113781435 A CN113781435 A CN 113781435A CN 202111062913 A CN202111062913 A CN 202111062913A CN 113781435 A CN113781435 A CN 113781435A
Authority
CN
China
Prior art keywords
network
defect
detection
yolov5
sample
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Withdrawn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
CN202111062913.2A
Other languages
Chinese (zh)
Inventor
谢小敏
马永骋
叶明�
刘凯
吴超
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.)
Changzhou Vocational Institute of Mechatronic Technology
Original Assignee
Changzhou Vocational Institute of Mechatronic Technology
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
Application filed by Changzhou Vocational Institute of Mechatronic Technology filed Critical Changzhou Vocational Institute of Mechatronic Technology
Priority to CN202111062913.2A priority Critical patent/CN113781435A/en
Publication of CN113781435A publication Critical patent/CN113781435A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • 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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a cigarette packet appearance defect detection method based on a YOLOV5 network, and belongs to the field of defect detection. The method specifically comprises the following steps: s1, collecting original images of the cigarette packet at four stations; s2, marking a defect sample in the collected original image to form a defect sample set; s3, enhancing and dividing the data of the defect sample set; s4, constructing a YOLOV5 network, and acquiring optimal network model parameters by using a transfer learning training network; and S5, completing sample prediction and carrying out subsequent actions. The invention applies the YOLOV5 network to the cigarette packet appearance defect detection, automatically identifies various appearance defects, and has high processing speed and high detection precision.

Description

Cigarette packet appearance defect detection method based on YOLOV5 network
Technical Field
The invention relates to a cigarette packet appearance defect detection method based on a YOLOV5 network, and belongs to the field of defect detection.
Background
Along with the development of science and technology, cigarette enterprises adopt intelligent equipment to carry out high-speed automatic packaging on cigarettes. In the process, defects such as false teeth, breakage, turning, deformation and the like of the pulling line of the small cigarette packet inevitably occur, so that the brand image of an enterprise is seriously influenced, the satisfaction degree of a consumer is reduced, and the competitiveness of the enterprise is negatively influenced. Therefore, many enterprises adopt a spot inspection method to monitor the cigarette packaging quality, and the method has randomness and is not beneficial to quality tracing. At present, cigarette appearance defect detection means adopted in the cigarette industry mainly rely on a traditional machine vision method, namely, a plurality of detection frames are manually arranged, a template of a qualified product is established according to design characteristics such as color, pattern or texture of cigarettes, a product to be detected is compared with the template, and the product is a defective product when the set threshold value is exceeded. The method can effectively realize the appearance detection of the cigarette packet, and both the recognition rate and the processing speed meet the production requirement. However, the method needs to manually set a detection frame, so that the method has the disadvantages of large dependence on experience, complex parameter adjustment interface, difficult operation of users, only processing the defect of a known position, and poor flexibility. In recent years, the method based on deep learning shows great superiority in the field of target detection and identification. The deep learning method does not need to manually design features, can automatically identify the appearance defect features of the small bags, and can effectively overcome the influence of factors such as illumination or shaking in transmission. In recent years, deep network models based on regional suggestions, such as R-CNN, Fast RCNN and the like, are successfully applied to industrial detection sites, but the processing speed of the network models is low, and the real-time requirement of high-speed transmission cigarette packet appearance detection cannot be met. Aiming at the problems, the invention applies the YOLOV5 network to the cigarette packet appearance defect detection, automatically identifies various appearance defects, and has high processing speed and high detection precision.
Disclosure of Invention
The invention aims to provide a cigarette packet appearance defect detection method based on a YOLOV5 network, which solves the problems that the existing detection method has low processing speed and cannot meet the real-time requirement of high-speed transmission cigarette packet appearance detection.
A cigarette packet appearance defect detection method based on a YOLOV5 network comprises the following steps:
s1, collecting original images of the cigarette packet at four stations;
s2, marking a defect sample in the collected original image to form a defect sample set;
s3, enhancing and dividing the data of the defect sample set;
s4, constructing a YOLOV5 network, and acquiring optimal network model parameters by using a transfer learning training network;
and S5, completing sample prediction and carrying out subsequent actions.
Further, in S1, specifically, when the cigarettes traveling on the high-speed conveying device reach the position of the photoelectric sensor, the industrial cameras at the four stations are triggered to operate, and the original images of the cigarette packets are collected from the front, the top view, the left side and the right side, wherein the size of the collected image is 640 × 480.
Further, in S2, specifically, a label img tool is used to mark each type of defect sample appearing in the appearance of the cigarette packet, position information of the defect is obtained, the type is set as defect type, and the sample amount of each type of defect sample for defect marking is not less than 300.
Further, the categories of defect samples include breakage, flipping, and deformation.
Further, in S3, specifically, the tobacco bale images are spliced by random flipping, random clipping, random scaling, and random arrangement to complete the expansion of the labeled data set, and the enhanced data set is divided into three types, i.e., a training set, a verification set, and a test set, according to a fixed ratio.
Further, in S4, the YOLOV5 network is divided into four models, YOLOV5S, YOLOV5m, YOLOV5l and YOLOV5x, according to the difference of the depth and width of the network, and the invention takes YOLOV5S as an example to construct a network model for detecting the appearance of the cigarette packet. And constructing a YOLOV5s network model containing four parts of an input end, a backhaul, a Neck and a Prediction. The input end adopts a K-means clustering method to obtain an initial anchor frame of the appearance defects of the cigarette packet according to the real marking frame of the defect sample; and the size of the image is processed by adopting a self-adaptive image scaling method, so that the defect detection speed is improved. The Backbone part adopts a Focus structure to slice the characteristic image, divides the characteristic diagram of the base layer into two parts, fuses the characteristic diagram through a CSP structure, combines richer gradient characteristics and reduces the calculation amount. The Neck part is mainly used for generating a characteristic pyramid, and the structures of the FPN and the PAN are adopted to enhance the fusion capability of network characteristics. The Prediction frame of the target is output by adopting a self-adaptive anchor frame method at the output end of the Prediction, the iterative network parameters are reversely updated according to the difference compared with the real frame, and the Bounding Box Loss calculation is carried out by adopting the GIOU _ Loss; setting the area ratio IOU of the overlapped area, adopting a non-maximum value inhibition method to inhibit redundant detection frames, namely sequencing all the detection frames according to scores, selecting the detection frame corresponding to the highest score, traversing other frames, deleting the detection frame if the IOU of the current highest score detection frame is larger than a set threshold value, continuously selecting one highest score frame from the other detection frames, repeating the process, finding all the reserved detection frames, and obtaining the final detection result.
Further, in S5, specifically, a network model of a large-scale match is used as an initial network parameter, the number of images and the number of iterations of each training are set, an optimal network parameter is obtained by training, an input sample is predicted by using the trained network parameter model, prediction results of images of four stations are obtained, a defect target is detected at any station, the sample is an unqualified sample, and the result is output to an execution mechanism for rejection.
The invention has the following beneficial effects: the invention relates to a cigarette packet appearance defect detection method based on a YOLOV5 network, which is characterized in that four station images of a cigarette packet are collected by an industrial camera, the detection of the cigarette packet appearance defects is completed by utilizing the YOLOV5 network, and the predicted category and position information is obtained, so that the subsequent mechanism processing is facilitated. By adopting the detection method, the defect target characteristics do not need to be designed manually, the detection frame and the parameters are set manually, various defects can be identified automatically, the method is simple, and the detection precision is high. The appearance defects of the cigarette packet are detected by adopting a YOLOV5 network, the processing speed is high, and the real-time requirement of an industrial field can be met. The detection algorithm of the invention can be suitable for detecting the appearance defects of the cigarette packets of different brands and has strong expansibility.
Drawings
Fig. 1 is a flow chart of a method for detecting appearance defects of cigarette packets based on a YOLOV5 network.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the invention provides a cigarette packet appearance defect detection method based on a YOLOV5 network, and the cigarette packet appearance defect detection method based on a YOLOV5 network comprises the following steps:
s1, collecting original images of the cigarette packet at four stations;
s2, marking a defect sample in the collected original image to form a defect sample set;
s3, enhancing and dividing the data of the defect sample set;
s4, constructing a YOLOV5 network, and acquiring optimal network model parameters by using a transfer learning training network;
and S5, completing sample prediction and carrying out subsequent actions.
Further, in S1, specifically, when the cigarettes traveling on the high-speed conveying device reach the position of the photoelectric sensor, the industrial cameras at the four stations are triggered to operate, and the original images of the cigarette packets are collected from the front, the top view, the left side and the right side, wherein the size of the collected image is 640 × 480.
Further, in S2, specifically, a label img tool is used to mark each type of defect sample appearing in the appearance of the cigarette packet, position information of the defect is obtained, the type is set as defect type, and the sample amount of each type of defect sample for defect marking is not less than 300.
The method for detecting appearance defects of cigarette packets based on the Yolov5 network as claimed in claim 3, wherein the categories of defect samples include breakage, turning over and deformation.
Further, in S3, specifically, the tobacco bale images are spliced by random flipping, random clipping, random scaling, and random arrangement to complete the expansion of the labeled data set, and the enhanced data set is divided into three types, i.e., a training set, a verification set, and a test set, according to a fixed ratio.
Further, in S4, a network model for detecting the appearance of the cigarette packet is constructed, specifically, a YOLOV5S network model including four major parts of an input end, a backhaul, a cock, and a preview is constructed.
Further, the input end obtains an initial anchor frame of the appearance defects of the cigarette packet by adopting a K-means clustering method according to a real marking frame of the defect sample, and the size of the image is processed by adopting a self-adaptive image scaling method;
the Backbone part adopts a Focus structure to slice the characteristic image, and divides the characteristic image of the basic layer into two parts;
the Neck part is mainly used for generating a characteristic pyramid, and the structures of the FPN and the PAN are adopted to enhance the fusion capability of network characteristics.
The Prediction frame of the target is output by adopting a self-adaptive anchor frame method at the output end of the Prediction, the iterative network parameters are reversely updated according to the difference compared with the real frame, and the Bounding Box Loss calculation is carried out by adopting the GIOU _ Loss; setting the area ratio IOU of the overlapped area, adopting a non-maximum value inhibition method to inhibit redundant detection frames, namely sequencing all the detection frames according to scores, selecting the detection frame corresponding to the highest score, traversing other frames, deleting the detection frame if the IOU of the current highest score detection frame is larger than a set threshold value, continuously selecting one highest score frame from the other detection frames, repeating the process, finding all the reserved detection frames, and obtaining the final detection result.
Further, in S5, specifically, a network model of a large-scale match is used as an initial network parameter, the number of images and the number of iterations of each training are set, an optimal network parameter is obtained by training, an input sample is predicted by using the trained network parameter model, prediction results of images of four stations are obtained, a defect target is detected at any station, the sample is an unqualified sample, and the result is output to an execution mechanism for rejection.
The invention discloses an appearance defect detection method based on a Yolov5 network, which can be applied to cigarette packet appearance defect detection and comprises the steps of collecting original images of four stations of a cigarette packet, carrying out defect marking work of unqualified samples, expanding a data set by using a data enhancement means, dividing the expanded data set into a training set, a verification set and a training set according to proportion, constructing a Yolov5 network model, and obtaining an optimal weight parameter of a Yolov5 network by using transfer learning; and predicting the input image by using the trained parameters to obtain the predicted category and position information, wherein the prediction result can be used as the basis for subsequent elimination and other processing. The invention utilizes the YOLOV5 network to detect the appearance defects of the cigarette packets, can automatically identify various defects, has simple method and high processing speed, can meet the real-time requirement of an industrial field, and can be popularized to cigarette packet brands.
The above embodiments are only used to help understanding the method of the present invention and the core idea thereof, and a person skilled in the art can also make several modifications and decorations on the specific embodiments and application scope according to the idea of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims (8)

1. A cigarette packet appearance defect detection method based on a YOLOV5 network is characterized by comprising the following steps of:
s1, collecting original images of the cigarette packet at four stations;
s2, marking a defect sample in the collected original image to form a defect sample set;
s3, enhancing and dividing the data of the defect sample set;
s4, constructing a YOLOV5 network, and acquiring optimal network model parameters by using a transfer learning training network;
and S5, completing sample prediction and carrying out subsequent actions.
2. The method for detecting the appearance defects of cigarette packets based on the YOLOV5 network as claimed in claim 1, wherein in S1, when the cigarettes running on the high-speed conveyor reach the position of the photoelectric sensor, the industrial cameras of four stations are triggered to operate, and the original images of the cigarette packets are collected from the front, the top, the left side and the right side, wherein the size of the collected images is 640 x 480.
3. The method for detecting the appearance defects of the cigarette packets based on the YOLOV5 network as claimed in claim 1, wherein in S2, specifically, a label img tool is used to mark various defect samples appearing in the appearance of the cigarette packets, position information of the defects is obtained, the types are set as defect types, and the sample amount of each type of defect sample for marking the defects is not less than 300.
4. The method for detecting appearance defects of cigarette packets based on the Yolov5 network as claimed in claim 3, wherein the categories of defect samples include breakage, turning over and deformation.
5. The method for detecting the appearance defects of the cigarette packets based on the YOLOV5 network as claimed in claim 1, wherein in S3, specifically, the cigarette packet images are spliced by random flipping, random cutting, random scaling and random arrangement to complete the expansion of the labeled data set, and the enhanced data set is divided into three types, namely a training set, a verification set and a test set according to a fixed proportion.
6. The method for detecting the appearance defects of the cigarette packets based on the YOLOV5 network as claimed in claim 1, wherein in S4, a network model for the appearance detection of the cigarette packets is constructed, specifically, a YOLOV5S network model containing four major parts of inputs, backsbone, Neck and Prediction is constructed.
7. The method for detecting the appearance defects of the cigarette packets based on the Yolov5 network as claimed in claim 6, wherein the input end obtains the initial anchor frame of the appearance defects of the cigarette packets by using a K-means clustering method according to the real mark frame of the defect samples, and the size of the image is processed by using an adaptive image scaling method;
the Backbone part adopts a Focus structure to slice the characteristic image, and divides the characteristic image of the basic layer into two parts;
the Neck part is mainly used for generating a characteristic pyramid, and the structures of the FPN and the PAN are adopted to enhance the fusion capability of network characteristics;
the Prediction frame of the target is output by adopting a self-adaptive anchor frame method at the output end of the Prediction, the iterative network parameters are reversely updated according to the difference compared with the real frame, and the Bounding Box Loss calculation is carried out by adopting the GIOU _ Loss; setting the area ratio IOU of the overlapped area, adopting a non-maximum value inhibition method to inhibit redundant detection frames, namely sequencing all the detection frames according to scores, selecting the detection frame corresponding to the highest score, traversing other frames, deleting the detection frame if the IOU of the current highest score detection frame is larger than a set threshold value, continuously selecting one highest score frame from the other detection frames, repeating the process, finding all the reserved detection frames, and obtaining the final detection result.
8. The method for detecting the appearance defects of the cigarette packets based on the YOLOV5 network as claimed in claim 1, wherein in S5, specifically, a network model of a large-scale competition is used as an initial network parameter, the number of images and the number of iterations of each training are set, the training obtains an optimal network parameter, an input sample is predicted by using the trained network parameter model, the prediction results of the images at four stations are obtained, a defect target is detected at any station, the sample is a non-qualified sample, and the result is output to an execution mechanism for removal.
CN202111062913.2A 2021-09-10 2021-09-10 Cigarette packet appearance defect detection method based on YOLOV5 network Withdrawn CN113781435A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111062913.2A CN113781435A (en) 2021-09-10 2021-09-10 Cigarette packet appearance defect detection method based on YOLOV5 network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111062913.2A CN113781435A (en) 2021-09-10 2021-09-10 Cigarette packet appearance defect detection method based on YOLOV5 network

Publications (1)

Publication Number Publication Date
CN113781435A true CN113781435A (en) 2021-12-10

Family

ID=78842567

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111062913.2A Withdrawn CN113781435A (en) 2021-09-10 2021-09-10 Cigarette packet appearance defect detection method based on YOLOV5 network

Country Status (1)

Country Link
CN (1) CN113781435A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114627126A (en) * 2022-05-17 2022-06-14 北京东方国信科技股份有限公司 Nuclear fuel rod defect detection method and device and nuclear reaction system
CN114677597A (en) * 2022-05-26 2022-06-28 武汉理工大学 Gear defect visual inspection method and system based on improved YOLOv5 network

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114627126A (en) * 2022-05-17 2022-06-14 北京东方国信科技股份有限公司 Nuclear fuel rod defect detection method and device and nuclear reaction system
CN114677597A (en) * 2022-05-26 2022-06-28 武汉理工大学 Gear defect visual inspection method and system based on improved YOLOv5 network
CN114677597B (en) * 2022-05-26 2022-10-11 武汉理工大学 Gear defect visual inspection method and system based on improved YOLOv5 network

Similar Documents

Publication Publication Date Title
CN110175982B (en) Defect detection method based on target detection
CN109961049B (en) Cigarette brand identification method under complex scene
CN105608456B (en) A kind of multi-direction Method for text detection based on full convolutional network
CN106960195B (en) Crowd counting method and device based on deep learning
CN108562589A (en) A method of magnetic circuit material surface defect is detected
CN112052797A (en) MaskRCNN-based video fire identification method and system
CN106845408A (en) A kind of street refuse recognition methods under complex environment
CN113781435A (en) Cigarette packet appearance defect detection method based on YOLOV5 network
CN107679502A (en) A kind of Population size estimation method based on the segmentation of deep learning image, semantic
CN114170161A (en) Cigarette appearance defect detection method and system
CN109544522A (en) A kind of Surface Defects in Steel Plate detection method and system
CN110598698B (en) Natural scene text detection method and system based on adaptive regional suggestion network
CN112053317A (en) Workpiece surface defect detection method based on cascade neural network
CN105718912A (en) Vehicle characteristic object detection method based on deep learning
CN115272204A (en) Bearing surface scratch detection method based on machine vision
CN112132196A (en) Cigarette case defect identification method combining deep learning and image processing
CN116052222A (en) Cattle face recognition method for naturally collecting cattle face image
CN115829995A (en) Cloth flaw detection method and system based on pixel-level multi-scale feature fusion
CN110276279A (en) A kind of arbitrary shape scene text detection method based on image segmentation
CN115816460A (en) Manipulator grabbing method based on deep learning target detection and image segmentation
CN112883755A (en) Smoking and calling detection method based on deep learning and behavior prior
CN110508510A (en) A kind of plastic pump defect inspection method, apparatus and system
CN111523453A (en) Intelligent road surface cleanliness judging method based on machine vision
CN117079125A (en) Kiwi fruit pollination flower identification method based on improved YOLOv5
CN116542962A (en) Improved Yolov5m model-based photovoltaic cell defect detection method

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20211210

WW01 Invention patent application withdrawn after publication