CN108960331A - A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering - Google Patents
A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering Download PDFInfo
- Publication number
- CN108960331A CN108960331A CN201810751316.2A CN201810751316A CN108960331A CN 108960331 A CN108960331 A CN 108960331A CN 201810751316 A CN201810751316 A CN 201810751316A CN 108960331 A CN108960331 A CN 108960331A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- feature
- cluster
- image
- vector
- 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
- 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
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of recognition methods again of the pedestrian based on pedestrian image feature clustering, major technique, which is contemplated that, carries out pedestrian image local shape factor using depth convolutional neural networks, obtained pedestrian's feature is subjected to cluster operation, cluster operation may make obtained pedestrian's feature to have certain Semantic, to lose the local Semantic of image when guaranteeing characteristic matching.There is biggish correlation to lead to bulk redundancy information to measurer in pedestrian image feature cluster after cluster, de-correlation technique or dimensionality reduction operation removal redundancy can be used.Minimum distance method automatic aligning pedestrian image is reused later, cost function of the different pedestrian's features after calculating cluster as entire neural network.The entire neural network of training, the similarity of the characteristic key pedestrian and pedestrian retrieval collection that are extracted later with neural network are ranked up to obtain the result that pedestrian identifies again according to similarity.
Description
Yu Zhongyong
Technical field
The present invention relates to pedestrian's identification technology again, the problem of belonging to image retrieval in computer vision field, and in particular to
A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering.
Background technique
Pedestrian identifies again refers to a given interested people, and pedestrian identifies that Re-ID is needed in other times again, other ground
Point, other cameras are specified out by personage again.For training set, test set, its very big feature is on no ID
Overlapping.
Identification technology is divided into two kinds to pedestrian again: based on extract pedestrian's feature pedestrian again recognition methods and be based on metric learning
Pedestrian's recognition methods again.Pedestrian knows three ranks of another characteristic point, bottom visual signature: as (such as color is straight for color characteristic again
Side figure), textural characteristics (such as LBP)；Middle layer semantic feature: such as pedestrian's hair style, types of garments；Further feature: it is primarily referred to as
The feature automatically extracted by deep neural network.Deep neural network by various measurements lose (contrastive loss,
Triplet loss, quadruplet loss etc.) learning characteristic in a manner of end to end, already take up leading position.
It is an object of the invention to propose a kind of pedestrian image feature clustering method based on deep learning model, depth is utilized
Degree learning model automatically extracts body local feature, then the feature that the body local feature extracted carries out semantically is gathered
Class realizes a kind of recognition methods again of the pedestrian based on pedestrian image feature clustering in conjunction with characteristics of human body's alignment schemes.
Summary of the invention
It is an object of the invention to propose a kind of pedestrian image feature clustering method based on deep learning model, one is provided
The kind higher self-aligning pedestrian of the pedestrian image feature recognition methods again based on semantization of accuracy rate.Particular technique is realized: one
The kind recognition methods again of the pedestrian based on pedestrian image feature clustering, detailed description are as follows:
Characteristic image is extracted using CNN such as resnet50 for every input picture.The last layer convolutional layer it is defeated
The feature extracted out as CNN.For N input pictures, available N number of 3D tensor T.For obtained each 3D
Tensor T, carries out cluster operation, and specific clustering algorithm can be used according to actual effect.The purpose of cluster operation be attempt to by
The component of 3D tensor T is divided into several disjoint subsets, so that resulting tensor has certain local Semantic.To N
A 3D tensor T, which carries out cluster operation, can be obtained N number of 3D tensor T ', pass through horizontal pond/convolution again for each 3D tensor T '
Operation, obtains the column vector Gn of N number of 3D tensor T '.Each component of gained column vector is the feature of pedestrian image, feature
Similarity of the Euclidean distance as two pictures calculates the similarity of two feature vectors as entire by minimum distance method
The loss function of CNN.It is constantly trained on training set, reduces loss in conjunction with ordering techniques again and finally obtain the row of better performances
People's identification model again.
Recognition methods has the following advantages that and effect pedestrian based on pedestrian image feature clustering relative to other technologies again:
(1) present invention has carried out the cluster operation of body local feature after extracting the local feature of human body, enhances
The semantization of body local feature, can be further improved pedestrian's recognition accuracy again.Pedestrian's feature of local semantization is again
By decorrelation or dimensionality reduction, bulk redundancy characteristic information is removed, the calculation amount of characteristic matching is substantially reduced.
(2) the body local feature of semantic level may not be in same dimension, semantic office after body local feature clustering
Portion's feature can not correspond, so the similarity of two pictures is calculated using critical path method (CPM), it can be in conjunction with ordering techniques again
Greatly improve the accuracy rate that pedestrian identifies again.
Detailed description of the invention
Fig. 1 is pedestrian's identifying system flow chart again
Fig. 2 is pedestrian's recognition methods algorithm block diagram again based on pedestrian image cluster in the present invention
Specific embodiment
The present invention is for providing a kind of recognition methods again of the pedestrian based on pedestrian image feature clustering.Main skill of the invention
Art design: pedestrian image local shape factor is carried out using depth convolutional neural networks, obtained pedestrian's feature is gathered
Generic operation, cluster operation may make obtained pedestrian's feature to have certain Semantic, pedestrian can be improved and identify again accurately
Rate.It reuses minimum distance method and calculates cost function of the different pedestrian's features as entire neural network after clustering.Training is whole
A neural network, the similarity of the characteristic key pedestrian and pedestrian retrieval collection that are extracted later with neural network, according to similarity
It is ranked up to obtain the result that pedestrian identifies again.To keep the purpose of the present invention, technical solution, effect clearer, clear, below
The present invention is described in more detail, and specific implementation step is as follows:
One, selected deep learning framework, such as selection pytorch deep learning frame, and configure and develop environment, prepare
Good pedestrian identification database again, Market1501, CHUK03 etc..
Two, selected depth network model, such as ResNet50 put up CNN model using selected deep learning frame, most
The feature that the output of later layer is extracted as CNN.The extracted characteristics of image of CNN is a 3D tensor.For N inputs
The available N number of 3D tensor T of image.
Three, using the feature that CNN is extracted as the data source of clustering algorithm, common clustering algorithm, such as K are analyzed
Mean cluster, mean shift clustering, greatest hope cluster of gauss hybrid models etc., select and design the poly- of suitable vector rank
Class algorithm carries out cluster operation to pedestrian's feature.N number of 3D vector T can be obtained after cluster operation with certain
The pedestrian image feature 3D vector T of local Semantic '.
Four, by the 3D vector T of cluster operation ' there is certain local Semantic, the component of a vector in cluster has larger
Correlation, contain more redundancy.For the information of redundancy, de-correlation technique such as orthogonalization or drop can be used
Dimension operation, it is ensured that integral operation amount is reduced in the case where accuracy rate.
Five, vector obtained in the previous step is subjected to horizontal pond or convolution operation, the column vector of available 3D vector.By
In the independence of cluster operation, the column vector dimension of every image may be different from.
Six, since pedestrian image feature is eventually converted to the inconsistent column vector of dimension, existing method can not use one
The similarity of different pedestrian images is calculated to one mode.The present invention carries out similarity meter using pedestrian's feature Auto-matching algorithm
Calculate, i.e. the algorithm Euclidean distance that calculates certain the smallest one-dimensional column vector of similarity automatically, by the Euclidean distance of institute's directed quantity and
Loss function as CNN.
Seven, on training set training deep neural network model, reduce loss, finally obtain better performances based on depth
The pedestrian of the pedestrian image feature clustering of study identification model again.
Claims (5)
1. a kind of recognition methods again of the pedestrian based on body local feature clustering, which comprises the following steps:
Step 1: selected deep learning framework, such as selection pytorch deep learning frame, and configure and develop environment, prepare
Good pedestrian identification database again, Market1501, CHUK03 etc..
Step 2: selected depth network model, such as ResNet50, CNN model is put up using selected deep learning frame, most
The feature that the output of later layer is extracted as CNN.The extracted characteristics of image of CNN is a 3D tensor.For N inputs
The available N number of 3D tensor T of image.
Step 3: analyzing common clustering algorithm, such as K using the feature that CNN is extracted as the data source of clustering algorithm
Mean cluster, mean shift clustering, greatest hope cluster of gauss hybrid models etc., select and design the poly- of suitable vector rank
Class algorithm carries out cluster operation to pedestrian's feature.N number of 3D vector T can be obtained after cluster operation with certain
The pedestrian image feature 3D vector T of local Semantic '.
Step 4: by the 3D vector T of cluster operation ' there is certain local Semantic, the component of a vector in cluster has larger
Correlation, contain more redundancy.For the information of redundancy, de-correlation technique such as orthogonalization or drop can be used
Dimension operation, it is ensured that integral operation amount is reduced in the case where accuracy rate.
Step 5: vector obtained in the previous step is carried out horizontal pond or convolution operation, the column vector of available 3D vector.By
In the independence of cluster operation, the column vector dimension of every image may be different from.
Step 6: existing method can not use one since pedestrian image feature is eventually converted to dimension inconsistent column vector
The similarity of different pedestrian images is calculated to one mode.The present invention carries out similarity meter using pedestrian's feature Auto-matching algorithm
Calculate, i.e. the algorithm Euclidean distance that calculates certain the smallest one-dimensional column vector of similarity automatically, by the Euclidean distance of institute's directed quantity and
Loss function as CNN.
Step 7: on training set training deep neural network model, reduce loss, finally obtain better performances based on depth
The pedestrian of the pedestrian image feature clustering of study identification model again.
2. the recognition methods again of the pedestrian based on body local feature clustering according to claim 1, which is characterized in that depth
The extracted feature of neural network needs to carry out cluster operation, automatic to carry out human part point by the automatic semantization of pedestrian's feature
It cuts.
3. the recognition methods again of the pedestrian based on body local feature clustering according to claim 1 or 2, which is characterized in that
After pedestrian image feature carries out cluster operation, convolution/pondization operation should be also carried out, redundancy is removed.
4. the recognition methods again of the pedestrian based on body local feature clustering according to claim 1,2 or 3, feature exist
In, due to semantization pedestrian image feature with pedestrian image feature to be retrieved is different surely corresponds, need to take automatic
Matching process is ignored level and is corresponded to, and the semantic feature of pedestrian image is used to carry out minimum distance calculation.
5. the recognition methods again of the pedestrian based on body local feature clustering according to claim 1 or 4, which is characterized in that
The similarity calculation of pedestrian image uses minimum distance method, and what it is in training stage calculating is two and its above pedestrian image
Similarity, test phase calculate be query set and Candidate Set pedestrian image similarity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810751316.2A CN108960331A (en) | 2018-07-10 | 2018-07-10 | A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810751316.2A CN108960331A (en) | 2018-07-10 | 2018-07-10 | A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108960331A true CN108960331A (en) | 2018-12-07 |
Family
ID=64483521
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810751316.2A Pending CN108960331A (en) | 2018-07-10 | 2018-07-10 | A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108960331A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110209658A (en) * | 2019-06-04 | 2019-09-06 | 北京字节跳动网络技术有限公司 | Data cleaning method and device |
| CN110427953A (en) * | 2019-06-21 | 2019-11-08 | 中南大学 | Robot is allowed to carry out the implementation method of vision place identification in changing environment based on convolutional neural networks and sequences match |
| CN110765954A (en) * | 2019-10-24 | 2020-02-07 | 浙江大华技术股份有限公司 | Vehicle weight recognition method, equipment and storage device |
| WO2020155606A1 (en) * | 2019-02-02 | 2020-08-06 | 深圳市商汤科技有限公司 | Facial recognition method and device, electronic equipment and storage medium |
| CN111639561A (en) * | 2020-05-17 | 2020-09-08 | 西北工业大学 | Unsupervised pedestrian re-identification method based on category self-adaptive clustering |
| CN111760292A (en) * | 2020-07-07 | 2020-10-13 | 网易（杭州）网络有限公司 | Method and device for detecting sampling data and electronic equipment |
| CN112560604A (en) * | 2020-12-04 | 2021-03-26 | 中南大学 | Pedestrian re-identification method based on local feature relationship fusion |
| CN112766180A (en) * | 2021-01-22 | 2021-05-07 | 重庆邮电大学 | Pedestrian re-identification method based on feature fusion and multi-core learning |
| CN113515995A (en) * | 2020-12-15 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Method, equipment and storage medium for re-recognition and model training of moving object |
| CN113723205A (en) * | 2021-08-04 | 2021-11-30 | 中国人民解放军陆军勤务学院 | Face recognition method based on face bottom library feature grouping |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915643A (en) * | 2015-05-26 | 2015-09-16 | 中山大学 | Deep-learning-based pedestrian re-identification method |
| CN106778464A (en) * | 2016-11-09 | 2017-05-31 | 深圳市深网视界科技有限公司 | A kind of pedestrian based on deep learning recognition methods and device again |
| CN106778604A (en) * | 2015-12-15 | 2017-05-31 | 西安电子科技大学 | Pedestrian's recognition methods again based on matching convolutional neural networks |
| CN106778527A (en) * | 2016-11-28 | 2017-05-31 | 中通服公众信息产业股份有限公司 | A kind of improved neutral net pedestrian recognition methods again based on triple losses |
| CN106897669A (en) * | 2017-01-19 | 2017-06-27 | 同济大学 | A kind of pedestrian based on consistent iteration various visual angles transfer learning discrimination method again |
| CN106980864A (en) * | 2017-03-31 | 2017-07-25 | 合肥工业大学 | A kind of pedestrian's recognition methods again based on support sample indirect type |
| CN107085609A (en) * | 2017-04-24 | 2017-08-22 | 国网湖北省电力公司荆州供电公司 | A kind of pedestrian retrieval method that multiple features fusion is carried out based on neutral net |
| CN107153817A (en) * | 2017-04-29 | 2017-09-12 | 深圳市深网视界科技有限公司 | Pedestrian's weight identification data mask method and device |
| CN107330396A (en) * | 2017-06-28 | 2017-11-07 | 华中科技大学 | A kind of pedestrian's recognition methods again based on many attributes and many strategy fusion study |
| CN107704890A (en) * | 2017-10-27 | 2018-02-16 | 北京旷视科技有限公司 | A kind of generation method and device of four-tuple image |
| AU2018100321A4 (en) * | 2018-03-15 | 2018-04-26 | Chen, Jinghan Mr | Person ReID method based on metric learning with hard mining |
| CN108171184A (en) * | 2018-01-03 | 2018-06-15 | 南京理工大学 | Method for distinguishing is known based on Siamese networks again for pedestrian |
-
2018
- 2018-07-10 CN CN201810751316.2A patent/CN108960331A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915643A (en) * | 2015-05-26 | 2015-09-16 | 中山大学 | Deep-learning-based pedestrian re-identification method |
| CN106778604A (en) * | 2015-12-15 | 2017-05-31 | 西安电子科技大学 | Pedestrian's recognition methods again based on matching convolutional neural networks |
| CN106778464A (en) * | 2016-11-09 | 2017-05-31 | 深圳市深网视界科技有限公司 | A kind of pedestrian based on deep learning recognition methods and device again |
| CN106778527A (en) * | 2016-11-28 | 2017-05-31 | 中通服公众信息产业股份有限公司 | A kind of improved neutral net pedestrian recognition methods again based on triple losses |
| CN106897669A (en) * | 2017-01-19 | 2017-06-27 | 同济大学 | A kind of pedestrian based on consistent iteration various visual angles transfer learning discrimination method again |
| CN106980864A (en) * | 2017-03-31 | 2017-07-25 | 合肥工业大学 | A kind of pedestrian's recognition methods again based on support sample indirect type |
| CN107085609A (en) * | 2017-04-24 | 2017-08-22 | 国网湖北省电力公司荆州供电公司 | A kind of pedestrian retrieval method that multiple features fusion is carried out based on neutral net |
| CN107153817A (en) * | 2017-04-29 | 2017-09-12 | 深圳市深网视界科技有限公司 | Pedestrian's weight identification data mask method and device |
| CN107330396A (en) * | 2017-06-28 | 2017-11-07 | 华中科技大学 | A kind of pedestrian's recognition methods again based on many attributes and many strategy fusion study |
| CN107704890A (en) * | 2017-10-27 | 2018-02-16 | 北京旷视科技有限公司 | A kind of generation method and device of four-tuple image |
| CN108171184A (en) * | 2018-01-03 | 2018-06-15 | 南京理工大学 | Method for distinguishing is known based on Siamese networks again for pedestrian |
| AU2018100321A4 (en) * | 2018-03-15 | 2018-04-26 | Chen, Jinghan Mr | Person ReID method based on metric learning with hard mining |
Non-Patent Citations (7)
| Title |
|---|
| LIN WU ET AL: "Deep adaptive feature embedding with local sample distributions for person re-identification", 《PATTERN RECOGNITION》 * |
| LIN WU ET AL: "What-and-where to match: Deep spatially multiplicative integration networks for person re-identification", 《PATTERN RECOGNITION》 * |
| XUAN ZHANG ET AL: "AlignedReID: Surpassing Human-Level Performance in Person Re-Identification", 《COMPUTER VISION AND PATTERN RECOGNITION 》 * |
| YIFAN SUN ET AL: "Beyond Part Models: Person Retrieval with Refined Part Pooling(and A Strong Convolutional Baseline)", 《COMPUTER VISION AND PATTERN RECOGNITION 》 * |
| ZONGTAO SONG ET AL: "Deep convolutional neural networks with adaptive spatial feature for person re-identification", 《2017 IEEE 2ND ADVANCED INFORMATION TECHNOLOGY,ELECTRONIC AND AUTOMATION CONTROL CONFERENCE(IAEAC) 》 * |
| 姚万超: "基于卷积神经网络的行人再识别算法", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 焦旭辉: "基于深度学习的行人再识别技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11455830B2 (en) | 2019-02-02 | 2022-09-27 | Shenzhen Sensetime Technology Co., Ltd. | Face recognition method and apparatus, electronic device, and storage medium |
| WO2020155606A1 (en) * | 2019-02-02 | 2020-08-06 | 深圳市商汤科技有限公司 | Facial recognition method and device, electronic equipment and storage medium |
| CN110209658A (en) * | 2019-06-04 | 2019-09-06 | 北京字节跳动网络技术有限公司 | Data cleaning method and device |
| CN110209658B (en) * | 2019-06-04 | 2021-09-14 | 北京字节跳动网络技术有限公司 | Data cleaning method and device |
| CN110427953A (en) * | 2019-06-21 | 2019-11-08 | 中南大学 | Robot is allowed to carry out the implementation method of vision place identification in changing environment based on convolutional neural networks and sequences match |
| CN110427953B (en) * | 2019-06-21 | 2022-11-29 | 中南大学 | Implementation method for enabling robot to perform visual place recognition in variable environment based on convolutional neural network and sequence matching |
| CN110765954A (en) * | 2019-10-24 | 2020-02-07 | 浙江大华技术股份有限公司 | Vehicle weight recognition method, equipment and storage device |
| CN111639561A (en) * | 2020-05-17 | 2020-09-08 | 西北工业大学 | Unsupervised pedestrian re-identification method based on category self-adaptive clustering |
| CN111760292A (en) * | 2020-07-07 | 2020-10-13 | 网易（杭州）网络有限公司 | Method and device for detecting sampling data and electronic equipment |
| CN112560604A (en) * | 2020-12-04 | 2021-03-26 | 中南大学 | Pedestrian re-identification method based on local feature relationship fusion |
| CN113515995A (en) * | 2020-12-15 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Method, equipment and storage medium for re-recognition and model training of moving object |
| CN112766180B (en) * | 2021-01-22 | 2022-07-12 | 重庆邮电大学 | Pedestrian re-identification method based on feature fusion and multi-core learning |
| CN112766180A (en) * | 2021-01-22 | 2021-05-07 | 重庆邮电大学 | Pedestrian re-identification method based on feature fusion and multi-core learning |
| CN113723205A (en) * | 2021-08-04 | 2021-11-30 | 中国人民解放军陆军勤务学院 | Face recognition method based on face bottom library feature grouping |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960331A (en) | A kind of recognition methods again of the pedestrian based on pedestrian image feature clustering | |
| CN110956185B (en) | Method for detecting image salient object | |
| CN109325952B (en) | Fashionable garment image segmentation method based on deep learning | |
| CN109784197B (en) | Pedestrian re-identification method based on hole convolution and attention mechanics learning mechanism | |
| CN108960059A (en) | A kind of video actions recognition methods and device | |
| CN110443818A (en) | A kind of Weakly supervised semantic segmentation method and system based on scribble | |
| CN113240691A (en) | Medical image segmentation method based on U-shaped network | |
| CN111429460A (en) | Image segmentation method, image segmentation model training method, device and storage medium | |
| CN104281572B (en) | A kind of target matching method and its system based on mutual information | |
| CN112070010B (en) | Pedestrian re-recognition method for enhancing local feature learning by combining multiple-loss dynamic training strategies | |
| CN110751027B (en) | Pedestrian re-identification method based on deep multi-instance learning | |
| CN104077742B (en) | Human face sketch synthetic method and system based on Gabor characteristic | |
| CN113033507B (en) | Scene recognition method and device, computer equipment and storage medium | |
| CN111985332B (en) | Gait recognition method of improved loss function based on deep learning | |
| CN113269224A (en) | Scene image classification method, system and storage medium | |
| CN109614933A (en) | A kind of motion segmentation method based on certainty fitting | |
| CN111582154A (en) | Pedestrian re-identification method based on multitask skeleton posture division component | |
| CN111881716A (en) | Pedestrian re-identification method based on multi-view-angle generation countermeasure network | |
| CN113221770A (en) | Cross-domain pedestrian re-identification method and system based on multi-feature hybrid learning | |
| Li et al. | Multi-label pattern image retrieval via attention mechanism driven graph convolutional network | |
| CN113870279A (en) | Multi-modal brain tumor image segmentation system and method | |
| CN117152459A (en) | Image detection method, device, computer readable medium and electronic equipment | |
| CN115984930A (en) | Micro expression recognition method and device and micro expression recognition model training method | |
| CN110427804A (en) | A kind of iris auth method based on secondary migration study | |
| CN114220082A (en) | Lane line identification method and device and computer 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: 20181207 |