CN111489334A - Defect workpiece image identification method based on convolution attention neural network - Google Patents

Defect workpiece image identification method based on convolution attention neural network Download PDF

Info

Publication number
CN111489334A
CN111489334A CN202010254533.8A CN202010254533A CN111489334A CN 111489334 A CN111489334 A CN 111489334A CN 202010254533 A CN202010254533 A CN 202010254533A CN 111489334 A CN111489334 A CN 111489334A
Authority
CN
China
Prior art keywords
attention
neural network
image
loss function
convolutional neural
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.)
Granted
Application number
CN202010254533.8A
Other languages
Chinese (zh)
Other versions
CN111489334B (en
Inventor
王永雄
蒋莉莉
刘智华
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.)
Warm House Information Technology Suzhou Co ltd
Original Assignee
Warm House Information Technology Suzhou Co ltd
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 Warm House Information Technology Suzhou Co ltd filed Critical Warm House Information Technology Suzhou Co ltd
Priority to CN202010254533.8A priority Critical patent/CN111489334B/en
Publication of CN111489334A publication Critical patent/CN111489334A/en
Application granted granted Critical
Publication of CN111489334B publication Critical patent/CN111489334B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • 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
    • G06T2207/30164—Workpiece; Machine component
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a defective workpiece image recognition method based on a convolutional attention neural network, which comprises the steps of pre-training weights on the convolutional neural network to form depth features which are used for inputting an input image by a feature extraction network; performing convolution operation on the depth characteristic image to obtain an attention diagram and performing normalization; performing attention area clipping and discarding on the normalized attention diagram, and inputting the normalized attention diagram into the convolutional neural network again for training; inputting a defect workpiece image to be identified into a feature map obtained by a trained convolutional neural network and performing dot product operation on the feature map and an attention map to obtain a new part feature map; and performing maximum pooling on the part feature map to obtain attention features, stacking all the attention features to obtain a part feature matrix, and completing classification and identification of the defect images. The invention improves the accuracy of workpiece defect detection and can be suitable for various micro defect detection tasks.

Description

Defect workpiece image identification method based on convolution attention neural network
Technical Field
The invention relates to a defective workpiece image identification method, in particular to a defective workpiece image identification method based on a convolution attention neural network.
Background
The casting production of the workpiece can have various defects, the defects on the surface of the workpiece can be judged by direct observation and other methods, but some defects exist in the workpiece, so that the defects cannot be judged by observation, and nondestructive detection is often carried out by means of X-ray images. Conventionally, the X-ray image is discriminated manually, and the discrimination efficiency is low. With the development of artificial intelligence, particularly the development of algorithms such as machine learning and the like, the efficiency is improved by applying a computer to judge defects. However, the machine learning algorithm also has the problem of poor universality, a corresponding model needs to be constructed for a workpiece, the system development is complicated, the efficiency is improved compared with that of manual identification on the whole, the identification process is still long in time consumption, and the accuracy is limited.
The problems of the existing non-contact nondestructive defect workpiece identification are mainly as follows: the defect image classification based on machine vision is a very challenging problem due to the diversity and subtle differences of workpiece defects, and a defect detection algorithm with good robustness and strong universality is designed, so that the effect is not ideal generally; defect identification based on a convolutional neural network has poor effect on tiny defects (such as crack defects) because the tiny defects occupy too small a space proportion of the whole image and are easily covered by information of other positions of the image.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a method for identifying the image of the defective workpiece based on the convolutional attention neural network, which solves the problem of low identification rate caused by the fact that tiny defects are easy to ignore.
The technical scheme of the invention is as follows: a defect workpiece image identification method based on a convolution attention neural network comprises the following steps:
step 1, pre-training weights are carried out on a convolutional neural network to form a feature extraction network, and an input image is input into the feature extraction network to obtain corresponding depth features;
step 2, performing convolution operation on the depth feature image formed by the depth features to obtain an attention diagram and performing normalization;
step 3, cutting an attention area in the normalized attention diagram, amplifying the area to the same size as the input image, and using the area as a new input of the convolutional neural network for training only in a classification task; discarding a part of attention areas in the attention diagram to obtain a new image which is used as a new input of the convolutional neural network and is only used for a classification task to train;
step 4, inputting the image of the defect workpiece to be identified into a depth feature map obtained by the trained convolutional neural network and performing dot product operation on the attention map obtained in the step 1 and the step 2 to obtain a new part feature map; and performing maximum pooling operation on the part feature map to obtain attention features, stacking all the attention features to obtain a part feature matrix, and completing classification and identification of the defect image.
Further, when the convolutional neural network is trained, a central loss function and a cross entropy loss function are used for weighted fusion to serve as a model optimization loss function.
Further, the central loss function
Figure BDA0002436777480000021
Wherein c iskAdding momentum β, c to the center of the site featurek←ck+β(fk-ck) To ckUpdate from 0, fkFor attention features, M is the dimension of the attention map, the model optimizes the loss function L-LCE+λLAWherein LCEλ is a weighting coefficient for the cross entropy loss function.
Further, when the convolutional neural network is trained by using the weighted fusion of the central loss function and the cross entropy loss function as the model optimization loss function, the convolutional neural network is trained by using a gradient descent method.
Further, the convolutional neural network is a VGG or ResNet network.
Further, the input image of step 1 is a workpiece data set with a defect category label.
The technical scheme provided by the invention has the advantages that:
on the basis of taking a Convolutional Neural Network (CNN) model as a workpiece classification network, an attention mechanism is introduced, important features in a feature map are extracted through a space attention method in the attention mechanism, then the feature map is multiplied by the feature map in the CNN network, the feature map is weighted, the weight of the important features in the feature map is increased, effective features are automatically screened, and the accuracy of workpiece defect detection is further improved.
And in the training process, data enhancement operation is carried out based on the attention diagram, and meanwhile, cutting and discarding operation is carried out on the data enhancement operation, so that the data set is further expanded, and more effective training sample data are generated. And a weighted fusion method of a central loss function and a cross entropy loss function is adopted to obtain more and better local or regional characteristics of the input image, and the accurate identification of fine-grained characteristics is ensured.
The convolutional attention neural network has wide application range and can be applied to various micro defect detection tasks, such as detection and identification of defects such as breakage and cracks on the surface of a ceramic tile.
Drawings
FIG. 1 is a schematic network flow diagram of a defective workpiece image identification method based on a convolutional attention neural network according to the present invention.
Fig. 2 is a schematic diagram illustrating an example of feature visualization of the convolutional neural network of the present invention.
Detailed Description
The present invention is further described in the following examples, which are intended to be illustrative only and not to be limiting as to the scope of the invention, which is to be given the full breadth of the appended claims and any and all equivalent modifications thereof which would occur to persons skilled in the art upon reading the present specification and which are intended to be within the scope of the present invention as defined in the appended claims.
Referring to fig. 1, a method for identifying a defective workpiece image based on a convolutional attention neural network according to this embodiment includes
Feature extraction, namely forming a feature extraction network by pre-training weights on a data set ImageNet based on a network structure of a convolutional neural network, and carrying out feature extraction on an image X ∈ R(H×W)Inputting the depth feature into a feature extraction network to obtain a corresponding depth feature;
the network structure of the convolutional neural network may be a VGG networkThe convolutional neural network adopted in the embodiment is ResNet34, which comprises 34 convolutional modules and is divided into 5 convolutional layers, and the depth feature of the output result of the fifth convolutional layer is reserved and is defined as a depth feature map F ∈ R(H×W×N)Wherein N represents the number of channels of the depth features, and W and H represent the width and height of each depth feature, respectively;
obtaining attention diagram, namely performing 1 × 1 convolution operation on the depth feature diagram, performing combined change on information of information among channels of the depth feature diagram, and obtaining an M-dimensional attention diagram defined as A ∈ R by using M1 × 1 convolution kernels(H×W×M)Wherein each fine-grained feature is defined as Ak∈R(H×W)(ii) a Normalizing the attention diagram to obtain an image A after normalizationk *＝(Ak-min(Ak))/(max(Ak)-min(Ak))；
Acquiring a cutting mask from the normalized attention map image, comparing each element in the attention map image, setting the pixel value larger than 0.5 as 1, and setting the rest as 0; obtaining a mask with local cutting characteristics, searching a marking frame capable of covering the whole mask area, amplifying the image in the whole marking frame to the size of an input image, and training as a new input; acquiring a discarding mask from the normalized attention image, and setting pixel values larger than 0.5 as 0 and setting the rest as 1 by comparing each element in the attention image; obtaining a mask image discarded from a part of attention parts, inputting the mask image as the latest input, retraining and learning more discriminative areas; new images generated after random cutting and random discarding only participate in classification, and do not participate in a new data enhancement process;
the obtained N-dimensional depth feature map F and the M-dimensional attention map a are subjected to bilinear pooling: multiplying two characteristics of each channel of the F and the A at the same position, and performing bilinear fusion to obtain a part characteristic diagram Fk＝Ak⊙ F (k is 1, 2.., M), and performing maximum pooling operation on the partial feature map to obtain the attention feature Fk∈R(1×N)(ii) a All will beStacking the M-dimensional attention features to obtain a part feature matrix P ∈ R(M×N)。
Through the steps, a plurality of regions with discriminant property can be obtained, and a central loss function is further introduced, so that each discriminant region is more diverse; defining a central loss function
Figure BDA0002436777480000031
Wherein c isk∈R(1×N)More diverse discriminative power regions are obtained by constraining the distance of each attention feature from the global center feature for the global feature center, thus by adding a momentum β, ck←ck+β(fk-ck) Updating the image from 0, calculating a model optimization loss function, calculating a central loss and a cross entropy loss function based on a position feature map, weighting and fusing to obtain more and better local or regional features of the input image and better solve the task of classifying fine-grained images, wherein the final model optimization loss function is L-LCE+λLAWherein LCEAnd (3) training the convolutional neural network by using a gradient descent method for a cross entropy loss function and lambda is a weighting coefficient.
Referring to fig. 2, the bright area in the maximum image in the figure is the detail feature of the object in the image, such as air hole, crack, etc., which is the focus of the model. The method comprehensively considers the characteristics of tiny defects in workpiece defect data set, the difference of different defect characteristics and other factors, and increases the difference among classes through an attention mechanism; and the loss function is optimized by combining with a model, the intra-class difference is reduced, and the characteristics with discriminant power and diversity are obtained, so that a good effect is obtained, the cost is saved compared with the existing method, and the actual deployment is easy. The convolution attention neural network is adopted, and the network is trained by combining the defective and non-defective workpiece image data sets labeled by categories, so that the convolution attention-based deep learning network capable of identifying the defective workpiece can be obtained. The trained network is used for detecting and identifying the defects of the workpiece, so that the accuracy and efficiency of detecting and identifying the defects of the workpiece can be improved, and the detection effect is improved.
The experimental result shows that the classification accuracy rate of the defect image can reach more than 96% in the test process. The method can be found out that the technology in the field of fine-grained image recognition is combined, so that the network can be helped to learn more areas with discrimination, and the classification accuracy is greatly improved.

Claims (6)

1. A defect workpiece image identification method based on a convolution attention neural network is characterized by comprising the following steps:
step 1, pre-training weights are carried out on a convolutional neural network to form a feature extraction network, and an input image is input into the feature extraction network to obtain corresponding depth features;
step 2, performing convolution operation on the depth feature image formed by the depth features to obtain an attention diagram and performing normalization;
step 3, cutting an attention area in the normalized attention diagram, amplifying the area to the same size as the input image, and using the area as a new input of the convolutional neural network for training only in a classification task; discarding a part of attention areas in the attention diagram to obtain a new image which is used as a new input of the convolutional neural network and is only used for a classification task to train;
step 4, inputting the image of the defect workpiece to be identified into the trained feature map obtained by the convolutional neural network and performing dot product operation on the attention map obtained in the step 1 and the step 2 to obtain a new part feature map; and performing maximum pooling operation on the part feature map to obtain attention features, stacking all the attention features to obtain a part feature matrix, and completing classification and identification of the defect image.
2. The method for identifying the image of the defective workpiece based on the convolutional attention neural network as claimed in claim 1, wherein the convolutional neural network is trained by using weighted fusion of a central loss function and a cross entropy loss function as a model optimization loss function.
3. The method of claim 2, wherein the central loss function is based on a defective workpiece image recognition method of a convolutional attention neural network
Figure FDA0002436777470000011
Wherein c iskAdding momentum β, c to the center of the site featurek←ck+β(fk-ck) To ckUpdate from 0, fkFor attention features, M is the dimension of the attention map, the model optimizes the loss function L-LCE+λLAWherein LCEλ is a weighting coefficient for the cross entropy loss function.
4. The method for identifying the image of the defective workpiece based on the convolutional attention neural network as claimed in claim 2, wherein when the convolutional neural network is trained by using weighted fusion of a central loss function and a cross entropy loss function as a model optimization loss function, the convolutional neural network is trained by using a gradient descent method.
5. The method of claim 1, wherein the convolutional neural network is a VGG or ResNet network.
6. The method of claim 1, wherein the input image of step 1 is a workpiece data set with a defect class label.
CN202010254533.8A 2020-04-02 2020-04-02 Defect workpiece image recognition method based on convolution attention neural network Active CN111489334B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010254533.8A CN111489334B (en) 2020-04-02 2020-04-02 Defect workpiece image recognition method based on convolution attention neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010254533.8A CN111489334B (en) 2020-04-02 2020-04-02 Defect workpiece image recognition method based on convolution attention neural network

Publications (2)

Publication Number Publication Date
CN111489334A true CN111489334A (en) 2020-08-04
CN111489334B CN111489334B (en) 2023-08-08

Family

ID=71812702

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010254533.8A Active CN111489334B (en) 2020-04-02 2020-04-02 Defect workpiece image recognition method based on convolution attention neural network

Country Status (1)

Country Link
CN (1) CN111489334B (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112200241A (en) * 2020-10-09 2021-01-08 山东大学 Automatic sorting method for fish varieties based on ResNet transfer learning
CN112381106A (en) * 2020-11-28 2021-02-19 南京理工大学 Target detection method based on global area prior attention
CN112699902A (en) * 2021-01-11 2021-04-23 福州大学 Fine-grained sensitive image detection method based on bilinear attention pooling mechanism
CN112801942A (en) * 2020-12-31 2021-05-14 广西慧云信息技术有限公司 Citrus huanglongbing image identification method based on attention mechanism
CN113034502A (en) * 2021-05-26 2021-06-25 深圳市勘察研究院有限公司 Drainage pipeline defect redundancy removing method
CN113362320A (en) * 2021-07-07 2021-09-07 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN113658183A (en) * 2021-10-21 2021-11-16 常州微亿智造科技有限公司 Workpiece quality inspection method and device and computer equipment
CN114677362A (en) * 2022-04-08 2022-06-28 四川大学 Surface defect detection method based on improved YOLOv5
CN115496945A (en) * 2022-09-19 2022-12-20 广东炬森智能装备有限公司 PCB true and false point identification and interpretable method based on normalized convolution attention mechanism
CN116091502A (en) * 2023-04-10 2023-05-09 厦门微图软件科技有限公司 Weak tag industrial defect image classification method based on improved depth residual error network
CN116385794A (en) * 2023-04-11 2023-07-04 河海大学 Robot inspection defect classification method and device based on attention flow transfer mutual distillation
CN116596875A (en) * 2023-05-11 2023-08-15 哈尔滨工业大学重庆研究院 Wafer defect detection method and device, electronic equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170124432A1 (en) * 2015-11-03 2017-05-04 Baidu Usa Llc Systems and methods for attention-based configurable convolutional neural networks (abc-cnn) for visual question answering
CN108345911A (en) * 2018-04-16 2018-07-31 东北大学 Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics
CN108830192A (en) * 2018-05-31 2018-11-16 珠海亿智电子科技有限公司 Vehicle and detection method of license plate under vehicle environment based on deep learning
CN109035233A (en) * 2018-07-24 2018-12-18 西安邮电大学 Visual attention network and Surface Flaw Detection method
CN109086779A (en) * 2018-07-28 2018-12-25 天津大学 A kind of attention target identification method based on convolutional neural networks

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170124432A1 (en) * 2015-11-03 2017-05-04 Baidu Usa Llc Systems and methods for attention-based configurable convolutional neural networks (abc-cnn) for visual question answering
CN108345911A (en) * 2018-04-16 2018-07-31 东北大学 Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics
CN108830192A (en) * 2018-05-31 2018-11-16 珠海亿智电子科技有限公司 Vehicle and detection method of license plate under vehicle environment based on deep learning
CN109035233A (en) * 2018-07-24 2018-12-18 西安邮电大学 Visual attention network and Surface Flaw Detection method
CN109086779A (en) * 2018-07-28 2018-12-25 天津大学 A kind of attention target identification method based on convolutional neural networks

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王一鸣,杜慧敏,张霞, 徐一丁: "视觉注意力网络在工件表面缺陷检测中的应用", 《计算机辅助设计与图形学学报》 *

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112200241A (en) * 2020-10-09 2021-01-08 山东大学 Automatic sorting method for fish varieties based on ResNet transfer learning
CN112381106B (en) * 2020-11-28 2022-09-09 南京理工大学 Target detection method based on global area prior attention
CN112381106A (en) * 2020-11-28 2021-02-19 南京理工大学 Target detection method based on global area prior attention
CN112801942A (en) * 2020-12-31 2021-05-14 广西慧云信息技术有限公司 Citrus huanglongbing image identification method based on attention mechanism
CN112801942B (en) * 2020-12-31 2023-10-13 广西慧云信息技术有限公司 Citrus yellow dragon disease image identification method based on attention mechanism
CN112699902A (en) * 2021-01-11 2021-04-23 福州大学 Fine-grained sensitive image detection method based on bilinear attention pooling mechanism
CN113034502A (en) * 2021-05-26 2021-06-25 深圳市勘察研究院有限公司 Drainage pipeline defect redundancy removing method
CN113362320A (en) * 2021-07-07 2021-09-07 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN113362320B (en) * 2021-07-07 2024-05-28 北京工业大学 Wafer surface defect mode detection method based on deep attention network
CN113658183A (en) * 2021-10-21 2021-11-16 常州微亿智造科技有限公司 Workpiece quality inspection method and device and computer equipment
CN114677362A (en) * 2022-04-08 2022-06-28 四川大学 Surface defect detection method based on improved YOLOv5
CN114677362B (en) * 2022-04-08 2023-09-12 四川大学 Surface defect detection method based on improved YOLOv5
CN115496945A (en) * 2022-09-19 2022-12-20 广东炬森智能装备有限公司 PCB true and false point identification and interpretable method based on normalized convolution attention mechanism
CN115496945B (en) * 2022-09-19 2023-04-18 广东炬森智能装备有限公司 PCB true and false point identification and interpretable method based on normalized convolution attention mechanism
CN116091502A (en) * 2023-04-10 2023-05-09 厦门微图软件科技有限公司 Weak tag industrial defect image classification method based on improved depth residual error network
CN116385794A (en) * 2023-04-11 2023-07-04 河海大学 Robot inspection defect classification method and device based on attention flow transfer mutual distillation
CN116385794B (en) * 2023-04-11 2024-04-05 河海大学 Robot inspection defect classification method and device based on attention flow transfer mutual distillation
CN116596875A (en) * 2023-05-11 2023-08-15 哈尔滨工业大学重庆研究院 Wafer defect detection method and device, electronic equipment and storage medium
CN116596875B (en) * 2023-05-11 2023-12-22 哈尔滨工业大学重庆研究院 Wafer defect detection method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
CN111489334B (en) 2023-08-08

Similar Documents

Publication Publication Date Title
CN111489334A (en) Defect workpiece image identification method based on convolution attention neural network
CN111179251B (en) Defect detection system and method based on twin neural network and by utilizing template comparison
CN109977808B (en) Wafer surface defect mode detection and analysis method
CN110175982B (en) Defect detection method based on target detection
CN109523518B (en) Tire X-ray defect detection method
CN112037219B (en) Metal surface defect detection method based on two-stage convolutional neural network
CN111860106B (en) Unsupervised bridge crack identification method
CN111540006B (en) Plant stomata intelligent detection and identification method and system based on deep migration learning
CN110717526A (en) Unsupervised transfer learning method based on graph convolution network
CN105574550A (en) Vehicle identification method and device
CN112241679B (en) Automatic garbage classification method
CN113378792B (en) Weak supervision cervical cell image analysis method fusing global and local information
CN111652273B (en) Deep learning-based RGB-D image classification method
CN113763312A (en) Detecting defects in semiconductor samples using weak marks
CN103106265A (en) Method and system of classifying similar images
CN112861635A (en) Fire and smoke real-time detection method based on deep learning
CN113221956B (en) Target identification method and device based on improved multi-scale depth model
CN112508857B (en) Aluminum product surface defect detection method based on improved Cascade R-CNN
CN113516650A (en) Circuit board hole plugging defect detection method and device based on deep learning
CN110929746A (en) Electronic file title positioning, extracting and classifying method based on deep neural network
CN113378883B (en) Fine-grained vehicle classification method based on channel grouping attention model
CN115439458A (en) Industrial image defect target detection algorithm based on depth map attention
CN110569780A (en) high-precision face recognition method based on deep transfer learning
CN116935092A (en) Automated defect classification and detection
CN111914902A (en) Traditional Chinese medicine identification and surface defect detection method based on deep neural network

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
GR01 Patent grant
GR01 Patent grant