CN110096947A - A kind of pedestrian based on deep learning recognizer again - Google Patents
A kind of pedestrian based on deep learning recognizer again Download PDFInfo
- Publication number
- CN110096947A CN110096947A CN201910198300.8A CN201910198300A CN110096947A CN 110096947 A CN110096947 A CN 110096947A CN 201910198300 A CN201910198300 A CN 201910198300A CN 110096947 A CN110096947 A CN 110096947A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- conv5
- neural network
- characteristic pattern
- training
- 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
- 238000013135 deep learning Methods 0.000 title claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 28
- 238000000034 method Methods 0.000 claims abstract description 12
- 239000012141 concentrate Substances 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims description 9
- 230000003321 amplification Effects 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 4
- 238000011840 criminal investigation Methods 0.000 description 2
- 102100040160 Rabankyrin-5 Human genes 0.000 description 1
- 101710086049 Rabankyrin-5 Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of pedestrian based on deep learning recognizer again, the method for the present invention are as follows: the neural network that Step1, building identify again for pedestrian；Step2, the neural network identified again for pedestrian is given using the parameter assignment of the training on ImageNet；Step3, using pedestrian, recognition training data set is trained the neural network identified again for pedestrian after assignment again；Step4, it is identified again using the image that the housebroken neural network identified again for pedestrian concentrates test data to carry out pedestrian.The present invention has more robust pedestrian's characteristic matching ability, can be efficiently used for pedestrian's identification, and the more conventionally known method of discrimination is high.
Description
Technical field
The present invention relates to a kind of pedestrian based on deep learning recognizers again, belong to field of image recognition.
Background technique
With the continuous maturation of pedestrian's identification technology again, this technology also starts to show its huge application value.?
Criminal investigation field, the demand that pedestrian's characteristic that recognizer models characteristics of human body again retrieves human body image with criminal investigation work
Mutually agree with.In retail domain, recognizer can help shopping mall supermarket operator to obtain effective customer track, identification to pedestrian again
Customer identification, to deeply excavate utilizable commercial value.In smart city, the successful implementation of corresponding intelligence system
Dependent on robust, high performance pedestrian recognizer again.
Feature extraction is carried out using convolutional neural networks and characteristic matching is two important sets of pedestrian's identification technology again
At part.Due to the scene complexity of the image from monitor video, existing difficulty needs overcome, for example, light, block,
The objective factors such as picture blur.In addition, the dress multiplicity of pedestrian, same people wear different clothes, different people wears similar clothing
Also to pedestrian, identification technology is put forward higher requirements clothes etc. again.The posture of pedestrian is changeable to lead to widely used alignment on face
Technology also fails in ReID.
Summary of the invention
The present invention provides a kind of pedestrian based on deep learning recognizers again.
The technical scheme is that a kind of pedestrian based on deep learning recognizer again, the specific step of the method
It is rapid as follows:
The neural network that Step1, building identify again for pedestrian；
Step2, the neural network identified again for pedestrian is given using the parameter assignment of the training on ImageNet；
Step3, using pedestrian again recognition training data set to the neural network identified again for pedestrian after assignment
It is trained；
Step4, it is carried out using the image that the housebroken neural network identified again for pedestrian concentrates test data
Pedestrian identifies again.
The parameter of the training on ImageNet are as follows: initial learning rate is 0.06, in 20 epoch and 40 epoch
When learning rate is decayed into 0.012 and 0.0024 respectively.
The training picture in recognition training data set using Random Level overturning does data amplification to the pedestrian again, by 60
A epoch training stops, and momentum is selected to be set as 0.9 SGD as optimizer.
The neural network identified again for pedestrian are as follows: using conv1, conv2_x of neural network resnet,
Conv3_x, conv4_x are as backbone network, using 4 branches similar with the conv5_x of resnet after conv4_x: the
One branch learns global characteristics, and second branch allows two parts of global characteristics transverse direction equal part to be learnt as local feature, the
Three branches allow three parts of global characteristics transverse direction equal part to be learnt as local feature, and the 4th branch makes global characteristics vertical etc.
It is divided to two parts to be learnt as local feature.
First branch meets the conv5_x of resnet, done respectively behind conv5_x global average pond and with it is complete
The maximum pond of office, then by the results added of the result in global average pond and overall situation maximum pond, then do 1 × 1 convolution operation,
By BN layers, relu layers, full articulamentum, again classified with softmax；
Second branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 2 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
The third branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 3 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
4th branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is vertically divided into 2 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x.
The beneficial effects of the present invention are: the present invention has more robust pedestrian's characteristic matching ability, can effectively use
It is identified in pedestrian, and the more conventionally known method of discrimination is high.
Detailed description of the invention
Fig. 1 is flow chart of the invention；
Fig. 2 is the conceptual schematic view of the neural network identified again in the present invention for pedestrian；
Fig. 3 is one schematic diagram of branch in the neural network structure identified again in the present invention for pedestrian；
Fig. 4 is two schematic diagram of branch in the neural network structure identified again in the present invention for pedestrian；
Fig. 5 is three schematic diagram of branch in the neural network structure identified again in the present invention for pedestrian；
Fig. 6 is four schematic diagram of branch in the neural network structure identified again in the present invention for pedestrian；
Fig. 7 is recognition result of embodiment of the present invention figure one；
Fig. 8 is recognition result of embodiment of the present invention figure two.
Specific embodiment
Embodiment 1: as shown in figures 1-8, a kind of pedestrian based on deep learning recognizer again, the specific step of the method
It is rapid as follows:
The neural network that Step1, building identify again for pedestrian；
Step2, the neural network identified again for pedestrian is given using the parameter assignment of the training on ImageNet；
Step3, using pedestrian again recognition training data set to the neural network identified again for pedestrian after assignment
It is trained；
Step4, it is carried out using the image that the housebroken neural network identified again for pedestrian concentrates test data
Pedestrian identifies again.
It is possible to further which the parameter of the training on ImageNet is arranged are as follows: initial learning rate is 0.06, at 20
Learning rate is decayed into 0.012 and 0.0024 respectively when epoch and 40 epoch.
It is possible to further which the pedestrian is arranged, the training picture in recognition training data set is overturn using Random Level again
Data amplification is done, is stopped by 60 epoch training, momentum is selected to be set as 0.9 SGD as optimizer.
It is possible to further which the neural network identified again for pedestrian is arranged are as follows: using neural network resnet's
Conv1, conv2_x, conv3_x, conv4_x are as backbone network, using the conv5_x phase with resnet after conv4_x
As 4 branches: first branch learns global characteristics, and second branch allows two parts of global characteristics transverse direction equal part as local spy
Sign is learnt, and third branch allows three parts of global characteristics transverse direction equal part to be learnt as local feature, and the 4th branch allows
Two parts of the vertical equal part of global characteristics is learnt as local feature.
It is possible to further which the conv5_x that first branch meets resnet is arranged, done respectively behind conv5_x
The average pond of the overall situation and with global maximum pond, then by the results added of the result in global average pond and overall situation maximum pond,
Then do 1 × 1 convolution operation, by BN layers, relu layers, full articulamentum, again classified with softmax；
Second branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 2 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
The third branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 3 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
4th branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is vertically divided into 2 parts of characteristic patterns
Horizontal stripe does global maximum pond behind every portion characteristic pattern, then does 1 × 1 convolution operation, by BN layers, relu layers, Quan Lian
It connects layer, classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x.
Specific simulation example is as follows:
Receive include pedestrian pedestrian image (such as Fig. 7, in 8 the leftmost side figure), in one embodiment of the present disclosure, row
People's image indicates the image including pedestrian to be identified；
The feature for extracting pedestrian image, extracts the process of feature specifically by the neural network, obtains from pedestrian image
To a three-dimensional tensor；Three-dimensional tensor passes through backbone network and 4 branches, finally obtains the feature of 4 branches.Hereafter, it incites somebody to action
To the features of four branches be stitched together by channel, picture identical with personage ID in picture is inputted is searched in search library
(in figure in addition to left side behind 10 pictures).
The figure of the leftmost side be it is to be measured attempt, after which is identified by means of the present invention, from search library go out examine
Preceding 10 picture (in figure in addition to left side behind 10 pictures) that rope goes out, 10 pictures obtained in Fig. 7 with it is to be tested
Pedestrian in figure is same people, have in 10 pictures obtained in Fig. 89 with it is to be measured attempt in pedestrian be same people (finally
One Zhang Butong), it is indicated above discrimination height.
After the method for the present invention and existing method (PCB, SVDNet) are compared, it is found that mAP value of the invention
It is higher, and preceding 1, preceding 5, preceding 10 discrimination is above existing method.As shown in table 1 below.
Table 1
| Method | mAP | rank 1 | rank 5 | rank10 |
| The present invention | 84.57 | 94.51 | 97.68 | 98.40 |
| PCB | 77.4 | 92.3 | 97.2 | 98.2 |
| SVDNet | 62.1 | 82.3 | 92.3 | 95.2 |
Above in conjunction with attached drawing, the embodiment of the present invention is explained in detail, but the present invention is not limited to above-mentioned
Embodiment within the knowledge of a person skilled in the art can also be before not departing from present inventive concept
It puts and makes a variety of changes.
Claims (5)
1. a kind of pedestrian based on deep learning recognizer again, it is characterised in that: specific step is as follows for the method:
The neural network that Step1, building identify again for pedestrian；
Step2, the neural network identified again for pedestrian is given using the parameter assignment of the training on ImageNet；
Step3, using pedestrian, recognition training data set carries out the neural network identified again for pedestrian after assignment again
Training；
Step4, pedestrian is carried out using the image that the housebroken neural network identified again for pedestrian concentrates test data
It identifies again.
2. the pedestrian according to claim 1 based on deep learning recognizer again, it is characterised in that: it is described
The parameter of training on ImageNet are as follows: initial learning rate is 0.06, in 20 epoch and 40 epoch respectively by learning rate
Decay to 0.012 and 0.0024.
3. the pedestrian according to claim 1 based on deep learning recognizer again, it is characterised in that: the pedestrian knows again
The training picture that other training data is concentrated does data amplification using Random Level overturning, stops by 60 epoch training, selects
Momentum is set as 0.9 SGD as optimizer.
4. the pedestrian according to claim 1 based on deep learning recognizer again, it is characterised in that: described to be used for pedestrian
The neural network identified again are as follows: using conv1, conv2_x, conv3_x, conv4_x of neural network resnet as backbone
Network, 4 branches similar with the conv5_x of resnet are used after conv4_x: first branch learns global characteristics, the
Two branches allow two parts of global characteristics transverse direction equal part to be learnt as local feature, and third branch makes global characteristics lateral etc.
Divide three parts to be learnt as local feature, the 4th branch allows two parts of the vertical equal part of global characteristics as local feature
It practises.
5. the pedestrian according to claim 4 based on deep learning recognizer again, it is characterised in that:
First branch meets the conv5_x of resnet, done respectively behind conv5_x global average pond and with the overall situation most
Great Chiization, then by the results added of the result in global average pond and global maximum pond, then do 1 × 1 convolution operation, pass through
BN layers, relu layers, full articulamentum, classified again with softmax；
Second branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 2 parts of characteristic pattern horizontal stripes,
Do global maximum pond behind every portion characteristic pattern, then do 1 × 1 convolution operation, by BN layers, relu layers, full articulamentum,
Classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
The third branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is laterally divided into 3 parts of characteristic pattern horizontal stripes,
Do global maximum pond behind every portion characteristic pattern, then do 1 × 1 convolution operation, by BN layers, relu layers, full articulamentum,
Classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x；
4th branch meets the conv5_x of resnet, and the characteristic pattern after conv5_x is vertically divided into 2 parts of characteristic pattern horizontal stripes,
Do global maximum pond behind every portion characteristic pattern, then do 1 × 1 convolution operation, by BN layers, relu layers, full articulamentum,
Classified again with softmax；Wherein, down-sampling is not used to operate in the conv5_1 in conv5_x.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910198300.8A CN110096947A (en) | 2019-03-15 | 2019-03-15 | A kind of pedestrian based on deep learning recognizer again |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910198300.8A CN110096947A (en) | 2019-03-15 | 2019-03-15 | A kind of pedestrian based on deep learning recognizer again |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110096947A true CN110096947A (en) | 2019-08-06 |
Family
ID=67443065
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910198300.8A Pending CN110096947A (en) | 2019-03-15 | 2019-03-15 | A kind of pedestrian based on deep learning recognizer again |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110096947A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765960A (en) * | 2019-10-29 | 2020-02-07 | 黄山学院 | Pedestrian re-identification method for adaptive multi-task deep learning |
| CN111177447A (en) * | 2019-12-26 | 2020-05-19 | 南京大学 | Pedestrian image identification method based on depth network model |
| CN111814857A (en) * | 2020-06-29 | 2020-10-23 | 浙江大华技术股份有限公司 | Target re-identification method, network training method thereof and related device |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107273836A (en) * | 2017-06-07 | 2017-10-20 | 深圳市深网视界科技有限公司 | A kind of pedestrian detection recognition methods, device, model and medium |
| CN107463881A (en) * | 2017-07-07 | 2017-12-12 | 中山大学 | A kind of character image searching method based on depth enhancing study |
| CN107704838A (en) * | 2017-10-19 | 2018-02-16 | 北京旷视科技有限公司 | The attribute recognition approach and device of destination object |
| CN108074211A (en) * | 2017-12-26 | 2018-05-25 | 浙江大华技术股份有限公司 | A kind of image processing apparatus and method |
| CN108090456A (en) * | 2017-12-27 | 2018-05-29 | 北京初速度科技有限公司 | A kind of Lane detection method and device |
| CN108491884A (en) * | 2018-03-27 | 2018-09-04 | 中山大学 | Pedestrian based on lightweight network identifying system and implementation method again |
| CN108764096A (en) * | 2018-05-21 | 2018-11-06 | 华中师范大学 | A kind of pedestrian weight identifying system and method |
| CN108875674A (en) * | 2018-06-29 | 2018-11-23 | 东南大学 | A kind of driving behavior recognition methods based on multiple row fusion convolutional neural networks |
| CN108960124A (en) * | 2018-06-28 | 2018-12-07 | 北京陌上花科技有限公司 | The image processing method and device identified again for pedestrian |
| CN108960127A (en) * | 2018-06-29 | 2018-12-07 | 厦门大学 | Pedestrian's recognition methods again is blocked based on the study of adaptive depth measure |
| CN109034044A (en) * | 2018-06-14 | 2018-12-18 | 天津师范大学 | A kind of pedestrian's recognition methods again based on fusion convolutional neural networks |
| CN109086672A (en) * | 2018-07-05 | 2018-12-25 | 襄阳矩子智能科技有限公司 | A kind of recognition methods again of the pedestrian based on reinforcement learning adaptive piecemeal |
| CN109087703A (en) * | 2018-08-24 | 2018-12-25 | 南京大学 | Abdominal cavity CT image peritonaeum metastatic marker method based on depth convolutional neural networks |
| US20190012170A1 (en) * | 2017-07-05 | 2019-01-10 | Deep Vision, Inc. | Deep vision processor |
| CN109359555A (en) * | 2018-09-21 | 2019-02-19 | 江苏安凰领御科技有限公司 | A kind of high-precision human face quick detection method |
-
2019
- 2019-03-15 CN CN201910198300.8A patent/CN110096947A/en active Pending
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107273836A (en) * | 2017-06-07 | 2017-10-20 | 深圳市深网视界科技有限公司 | A kind of pedestrian detection recognition methods, device, model and medium |
| US20190012170A1 (en) * | 2017-07-05 | 2019-01-10 | Deep Vision, Inc. | Deep vision processor |
| CN107463881A (en) * | 2017-07-07 | 2017-12-12 | 中山大学 | A kind of character image searching method based on depth enhancing study |
| CN107704838A (en) * | 2017-10-19 | 2018-02-16 | 北京旷视科技有限公司 | The attribute recognition approach and device of destination object |
| CN108074211A (en) * | 2017-12-26 | 2018-05-25 | 浙江大华技术股份有限公司 | A kind of image processing apparatus and method |
| CN108090456A (en) * | 2017-12-27 | 2018-05-29 | 北京初速度科技有限公司 | A kind of Lane detection method and device |
| CN108491884A (en) * | 2018-03-27 | 2018-09-04 | 中山大学 | Pedestrian based on lightweight network identifying system and implementation method again |
| CN108764096A (en) * | 2018-05-21 | 2018-11-06 | 华中师范大学 | A kind of pedestrian weight identifying system and method |
| CN109034044A (en) * | 2018-06-14 | 2018-12-18 | 天津师范大学 | A kind of pedestrian's recognition methods again based on fusion convolutional neural networks |
| CN108960124A (en) * | 2018-06-28 | 2018-12-07 | 北京陌上花科技有限公司 | The image processing method and device identified again for pedestrian |
| CN108875674A (en) * | 2018-06-29 | 2018-11-23 | 东南大学 | A kind of driving behavior recognition methods based on multiple row fusion convolutional neural networks |
| CN108960127A (en) * | 2018-06-29 | 2018-12-07 | 厦门大学 | Pedestrian's recognition methods again is blocked based on the study of adaptive depth measure |
| CN109086672A (en) * | 2018-07-05 | 2018-12-25 | 襄阳矩子智能科技有限公司 | A kind of recognition methods again of the pedestrian based on reinforcement learning adaptive piecemeal |
| CN109087703A (en) * | 2018-08-24 | 2018-12-25 | 南京大学 | Abdominal cavity CT image peritonaeum metastatic marker method based on depth convolutional neural networks |
| CN109359555A (en) * | 2018-09-21 | 2019-02-19 | 江苏安凰领御科技有限公司 | A kind of high-precision human face quick detection method |
Non-Patent Citations (5)
| Title |
|---|
| JIAWEI LIU 等: "CA3Net: Contextual-Attentional Attribute-Appearance Network for Person Re-Identification", 《AIXIV:1811.07544V1》 * |
| LIANG ZHENG 等: "Person Re-identification:Past, Present and Future", 《ARXIV:1610.02984V1》 * |
| RICCARDO MAZZON 等: "Person re-identification in crowd", 《ELSEVIER:PATTERN RECONGNITION LETTERS》 * |
| YIFAN SUN 等: "Beyond Part Models: Person Retrieval with Refined Part Pooling", 《ARXIV:1711.09349V1》 * |
| 徐阳: "基于卷积神经网络的行人重识别算法研究", 《中国优秀硕士论文全文数据库 信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765960A (en) * | 2019-10-29 | 2020-02-07 | 黄山学院 | Pedestrian re-identification method for adaptive multi-task deep learning |
| CN110765960B (en) * | 2019-10-29 | 2022-03-04 | 黄山学院 | Pedestrian re-identification method for adaptive multi-task deep learning |
| CN111177447A (en) * | 2019-12-26 | 2020-05-19 | 南京大学 | Pedestrian image identification method based on depth network model |
| CN111177447B (en) * | 2019-12-26 | 2021-04-30 | 南京大学 | Pedestrian image identification method based on depth network model |
| CN111814857A (en) * | 2020-06-29 | 2020-10-23 | 浙江大华技术股份有限公司 | Target re-identification method, network training method thereof and related device |
| CN111814857B (en) * | 2020-06-29 | 2021-07-06 | 浙江大华技术股份有限公司 | Target re-identification method, network training method thereof and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105956560B (en) | A kind of model recognizing method based on the multiple dimensioned depth convolution feature of pondization | |
| CN106778604B (en) | Pedestrian re-identification method based on matching convolutional neural network | |
| CN110096947A (en) | A kind of pedestrian based on deep learning recognizer again | |
| CN108764065A (en) | A kind of method of pedestrian's weight identification feature fusion assisted learning | |
| Sun et al. | Activity recognition using dense long-duration trajectories | |
| CN108960059A (en) | A kind of video actions recognition methods and device | |
| CN108921846A (en) | A kind of rail tread defect identification method combined based on gray level image and depth image | |
| CN106778856A (en) | A kind of object identification method and device | |
| Yao et al. | Robust CNN-based gait verification and identification using skeleton gait energy image | |
| CN109117897A (en) | Image processing method, device and readable storage medium storing program for executing based on convolutional neural networks | |
| CN105205453B (en) | Human eye detection and localization method based on depth self-encoding encoder | |
| CN105095884B (en) | A kind of pedestrian's identifying system and processing method based on random forest support vector machines | |
| CN105512618B (en) | Video tracing method | |
| CN105869178A (en) | Method for unsupervised segmentation of complex targets from dynamic scene based on multi-scale combination feature convex optimization | |
| CN109919977A (en) | A kind of video motion personage tracking and personal identification method based on temporal characteristics | |
| CN108805078A (en) | Video pedestrian based on pedestrian's average state recognition methods and system again | |
| CN111027377B (en) | Double-flow neural network time sequence action positioning method | |
| CN104751466B (en) | A kind of changing object tracking and its system based on conspicuousness | |
| CN113521749B (en) | Abnormal account detection model training method and abnormal account detection method | |
| CN103955950B (en) | Image tracking method utilizing key point feature matching | |
| CN104778472B (en) | Human face expression feature extracting method | |
| CN106295532A (en) | A kind of human motion recognition method in video image | |
| CN104751111A (en) | Method and system for recognizing human action in video | |
| CN116052218B (en) | Pedestrian re-identification method | |
| CN109635726A (en) | A kind of landslide identification method based on the symmetrical multiple dimensioned pond of depth network integration |
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 | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230714 |
|
| AD01 | Patent right deemed abandoned |