CN108921051B - Pedestrian attribute identification network and technology based on cyclic neural network attention model - Google Patents
Pedestrian attribute identification network and technology based on cyclic neural network attention model Download PDFInfo
- Publication number
- CN108921051B CN108921051B CN201810616398.XA CN201810616398A CN108921051B CN 108921051 B CN108921051 B CN 108921051B CN 201810616398 A CN201810616398 A CN 201810616398A CN 108921051 B CN108921051 B CN 108921051B
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- attribute
- neural network
- attributes
- network
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
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)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a pedestrian attribute identification network based on a recurrent neural network attention model and a pedestrian attribute identification technology. The pedestrian attribute identification network comprises a first convolution neural network for extracting the characteristics of a pedestrian whole-body image by using a pedestrian original whole-body image as an input; using the pedestrian whole-body image feature as a first input, using the attention heat map of the attribute group concerned at the previous moment as a second input, and outputting the attention heat map of the attribute group concerned at the current moment and a recurrent neural network passing through the partially highlighted pedestrian feature; and outputting a second convolutional neural network of the attribute prediction probability of the current attention group by using the pedestrian feature subjected to local highlight as an input. The invention utilizes the convolution cyclic neural network attention model to excavate the incidence relation of the spatial positions of the pedestrian attribute regions, more accurately highlights the positions of the regions corresponding to the attributes in the image, and realizes higher pedestrian attribute identification precision.
Description
Technical Field
The invention belongs to the technical field of neural networks and image recognition, and particularly relates to a pedestrian attribute recognition network and technology based on a recurrent neural network attention model.
Background
Pedestrian attribute identification technology can help people to automatically complete the task of searching for specific people from massive amounts of image and video data. However, due to the influence of factors such as low image quality of the surveillance video, small labeled pedestrian attribute data set, difficult acquisition and the like, the difficulty of identifying the pedestrian attributes from the surveillance video image is greatly increased. The existing pedestrian attribute identification method based on the deep neural network is divided into two categories, namely a Convolutional Neural Network (CNN) method and a convolutional neural network and cyclic neural network combined method (CNN-RNN). The existing CNN method such as the deep mar method attempts to identify the attribute of each pedestrian in isolation from the features of the entire image, and although this method achieves a certain effect, it ignores the correlation between the spatial locality of the attribute of the pedestrian and the attribute, and is difficult to obtain higher identification accuracy. The existing CNN-RNN method such as the JRL method tries to gradually mine the semantic association relation between pedestrian attributes by using a recurrent neural network, for example, a woman wearing a skirt is generally worn, and the recognition precision is improved to a certain extent compared with the pure CNN method. However, this approach only considers semantic links between pedestrian attributes but ignores the spatial locality of the attributes. The focus of many attributes of the pedestrian is concentrated in one area of the image, for example, whether the pedestrian wears glasses or leaves long hair is only determined by the visual characteristics of the head area of the pedestrian, and other areas are not useful. If the locality of the space is considered in the construction process of the pedestrian attribute identification model, the head area is highlighted when the head attribute is identified, and the interference of background noise is ignored, so that the pedestrian attribute identification precision can be greatly improved.
Disclosure of Invention
In order to solve the above technical problem, the present invention provides a pedestrian attribute identification network based on a recurrent neural network attention model, including:
a first convolution neural network for extracting the pedestrian whole-body image characteristics N (x) by using the pedestrian original whole-body image as input;
attention heat map A of attribute group concerned at last moment using pedestrian whole-body image feature N (x) as first inputt-1(x) As a second input, an attention heat map A of the set of attributes of interest at the current time is outputt(x) And the pedestrian characteristic H passing through local highlightt(x) A recurrent neural network of (a);
using partially highlighted pedestrian features Ht(x) As an input, a second convolutional neural network of attribute prediction probabilities for the current group of interest is output.
Further, the partially highlighted pedestrian feature Ht(x) Attention heat map A using the set of attributes that were focused on at the previous timet-1(x) The calculation formula is obtained by acting on the pedestrian whole-body image characteristic N (x) as follows:
further, a batch regularization operation is applied to the attribute prediction probability output to combat recognition errors caused by imbalance of the positive and negative example samples of the attributes.
Further, the pedestrian attribute identification network includes:
for each different attribute group of the same original pedestrian whole-body image, the state of a memory unit of the recurrent neural network is determined by the characteristics of the locally highlighted pedestrians of all the predicted attribute groups;
sharing the weight value of the first convolution neural network at different prediction moments;
the second convolutional neural network shares the weights for different predicted times.
Further, the pedestrian attribute identification network is trained by using a weighted Sigmoid cross entropy loss function, wherein the loss function is as follows:
wf＝exp(pj)
in the above formula, pjRepresenting the proportion of the positive case number of attribute j in the training set, wjThe learning weights representing the positive examples of the sample,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe label is the label of the jth attribute of the ith sample, N is the total number of training samples, and K is the total number of attributes to be identified.
The invention also provides a pedestrian attribute identification technology based on the recurrent neural network attention model, which comprises the following steps:
s1, acquiring a certain number of pedestrian images with attributes to be identified, marking whether the images have certain attributes or not, and acquiring a data set which can be used for training the identification effect of the attributes of the pedestrians; grouping all the labeled attributes according to semantic and spatial neighbor relations;
s2, combining an increment network with a convolution cyclic neural network to construct a pedestrian attribute identification network based on a convolution cyclic neural network attention model;
s3, defining a loss function required by training the pedestrian attribute identification network, and training the pedestrian attribute identification network constructed in the step S2 by using the training data set obtained in the step S1;
and S4, identifying the attributes in the pedestrian image to be identified by using the pedestrian attribute identification network trained in the step S3.
Further, the step S2 includes:
s2-1, extracting the original pedestrian whole-body image by using an inclusion network to obtain the pedestrian whole-body image characteristic N (x);
s2-2, at the moment i, calculating an attention heat map A of the attribute group concerned at the current moment by using the feature N (x) of the whole-body image of the pedestrian and a convolution cyclic neural networkt(x) And storing the historical information in a memory unit of the convolution cyclic neural network;
s2-3, use attention heatmap At(x) Acting on the pedestrian whole-body image feature N (x) to obtain the partially highlighted pedestrian feature Ht(x) The calculation formula is as follows:
s2-4, using the locally highlighted feature Ht(x) And carrying out attribute identification on the t-th group of attributes and outputting the prediction probability of the group of attributes.
Further, the loss function defined in step S3 is as follows:
wj＝exp(pj)
in the above formula, pjThe proportion of positive cases representing the attribute j in the training set, wjRepresentative examplesThe learning weight of the sample is determined,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe label is the label of the jth attribute of the ith sample, N is the total number of training samples, and K is the total number of attributes to be identified.
Compared with the prior art, the invention has the beneficial effects that:
the invention utilizes the convolution cyclic neural network attention model to excavate the incidence relation of the spatial positions of the pedestrian attribute regions, more accurately highlights the positions of the regions corresponding to the attributes in the image, and realizes higher pedestrian attribute identification precision.
Drawings
Fig. 1 is a structural diagram of a pedestrian attribute identification network based on a recurrent neural network attention model.
Detailed Description
Example 1
A pedestrian attribute identification network based on a recurrent neural network attention model, as shown in fig. 1, comprising:
a first convolution neural network for extracting the pedestrian whole-body image characteristics N (x) by using the pedestrian original whole-body image as input;
attention heat map A of attribute group concerned at last moment using pedestrian whole-body image feature N (x) as first inputt-1(x) As a second input, an attention heat map A of the set of attributes of interest at the current time is outputt(x) And pedestrian feature H passing through local highlightt(x) A recurrent neural network of (a);
using partially highlighted pedestrian features Ht(x) As an input, a second convolutional neural network of attribute prediction probabilities for the current group of interest is output.
In the pedestrian attribute identification network provided by the embodiment, the partially highlighted pedestrian feature Ht(x) Attention heat map A using the set of attributes that were focused on at the previous timet-1(x) The calculation formula is obtained by acting on the pedestrian whole-body image characteristic N (x) as follows:
in the pedestrian attribute identification network provided by this embodiment, a batch regularization operation is used for the attribute prediction probability output to counter identification errors caused by imbalance of the positive and negative example samples of the attributes.
In the pedestrian attribute identification network provided in this embodiment, the method further includes:
for each different attribute group of the same original pedestrian whole-body image, the state of a memory unit of the recurrent neural network is determined by the characteristics of the locally highlighted pedestrians of all the predicted attribute groups;
sharing the weight value of the first convolution neural network at different prediction moments;
the second convolutional neural networks share weights for different prediction instants.
In the pedestrian attribute recognition network provided in this embodiment, the pedestrian attribute recognition network is trained by using a weighted Sigmoid cross entropy loss function, where the loss function is as follows:
wj＝exp(pj)
in the above formula, pjRepresenting the proportion of the positive case number of attribute j in the training set, wjThe learning weights representing the positive examples of the sample,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe number of the labels of the jth attribute of the ith sample is N, the total number of the training samples is N, and the total number of the attributes to be identified is K.
Example 2
A pedestrian attribute identification technology based on a recurrent neural network attention model comprises the following steps:
s1, acquiring a certain number of pedestrian images with attributes to be identified, marking whether the images have certain attributes or not, and acquiring a data set which can be used for training the identification effect of the attributes of the pedestrians; then screening all the marked attributes, and grouping the attributes obtained by screening according to the semantic and spatial neighbor relations;
s2, combining an increment network with a convolution cyclic neural network to construct a pedestrian attribute identification network based on a convolution cyclic neural network attention model, and specifically comprising the following steps:
s2-1, extracting the original pedestrian whole-body image by using an inclusion network to obtain the pedestrian whole-body image characteristic N (x);
s2-2, at the moment i, calculating an attention heat map A of the attribute group concerned at the current moment by using the feature N (x) of the whole-body image of the pedestrian and a convolution cyclic neural networkt(x) And storing the historical information in a memory unit of the convolution cyclic neural network;
s2-3, use attention heatmap At(x) Acting on the pedestrian whole-body image feature N (x) to obtain the partially highlighted pedestrian feature Ht(x) The calculation formula is as follows:
s2-4, using the local highlighted feature Ht(x) Carrying out attribute identification on the t-th group of attributes and outputting the prediction probability of the group of attributes;
s3, defining a loss function required by training a pedestrian attribute recognition network, wherein the loss function is as follows:
wj＝exp(pj)
in the above formula, pjThe proportion of positive cases representing the attribute j in the training set, wjRepresents positiveThe learning weights of the sample are instantiated,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe attribute is a label of the jth attribute of the ith sample, N is the total number of training samples, and K is the total number of attributes to be identified;
training the pedestrian attribute recognition network constructed in step S2 using the training data set acquired in step S1; simultaneously, testing the pedestrian attribute recognition network obtained by training by using a test set;
and S4, identifying the attributes in the pedestrian image to be identified in the actual application scene by using the pedestrian attribute identification network obtained by training in the step S3.
The pedestrian attribute identification technique provided by the present invention is described in detail below on the basis of a pedestrian attribute identification RAP data set.
(1) And taking the pedestrian attribute recognition RAP data set as a data set used for training and testing the pedestrian attribute recognition effect. The RAP data set is a pedestrian attribute data set which is obtained by a team of a Chinese academy of Automation, 26 cameras are used for carrying out image acquisition on pedestrian monitoring videos in a mall, and 41,585 pedestrian images are finally screened out and added into the data set through analysis on context information of pedestrian attributes and environmental factors; and each image is labeled with 72 attributes including viewing angle information, presence or absence of occlusion, body part information, and the like.
(2) 72 attributes in the RAP data set are screened, 51 attributes which need to be used are screened, and the attributes are divided into 10 groups according to semantic and spatial neighbor relations, which is specifically shown in table 1.
TABLE 1 51 attributes in RAP dataset and corresponding groups
(3) Constructing a pedestrian attribute identification network shown in fig. 1, wherein the network utilizes a convolution cyclic neural network to train different grouping downlink attribute attention models, and utilizes an attention model combined with an increment convolution neural network to identify pedestrian attributes.
(4) Calculating the proportion p of each attribute label in the training set, wherein the positive sample accounts for all samplesj。
(5) Defining a loss function required for training a pedestrian attribute recognition network, and calculating p in the step (4)jThe substitution calculation is specifically as follows:
wj＝exp(pj)
(6) the random gradient descent algorithm is used for identifying the network of the pedestrian attributes, and the hyper-parameter setting of the training process is as follows:
initial learning rate: 0.1, Batch Size (Batch Size): and 64, lowering the learning rate to 1/10 of the initial learning rate every 10000 rounds, and using the depth model which is pre-trained on the Imagenet image classification task as an initial value of the pedestrian attribute identification model.
(7) And (3) under an actual test scene, inputting the image to be detected into the pedestrian attribute identification network obtained by training in the step (6), and outputting the prediction probability vectors corresponding to the grouping attributes in the step (2) by the network for 10 times, wherein the number of the prediction probability vectors is 51. And for the probability output corresponding to each attribute, if the probability value is greater than 0.5, the attribute is considered to be possessed, otherwise, the attribute is not considered to be possessed. And sequentially judging the probability output of each attribute, and finally outputting an identification result of all 51 attributes of the pedestrian.
Compared with the existing pedestrian attribute identification method, the pedestrian attribute identification technology based on the recurrent neural network attention model has higher identification precision. The pedestrian attribute identification technology provided by the invention evaluates two mainstream pedestrian attribute identification public data sets at present, and obtains higher evaluation precision than the conventional CNN method and the CNN-RNN method.
For the pedestrian attribute identification precision, mA (mean accuracy) is generally adopted to measure the quality of an attribute identification algorithm, and due to the characteristic of unbalanced attribute distribution, in order to ensure the reasonability of an accuracy calculation result, the mA respectively calculates the accuracy of a positive case and the accuracy of a negative case aiming at each attribute, takes an average value as the accuracy of attribute identification, and then integrates the average values of the accuracyands of all the attributes to calculate the final mA value of the attribute. The mA is calculated as follows:
wherein L represents the number of attributes; p isiRepresents the number of positive examples, TPiNumber of positive cases representing correct predictions; n is a radical ofiNumber of negative examples, TNiRepresenting the number of negative cases of correct prediction.
Compared with the DeepMAR method provided by the background art, the mA value of the pedestrian attribute identification technology provided by the invention is increased by 8.76%, and compared with the JRL method, the mA value is increased by 3.35%. In addition, the pedestrian attribute recognition technology provided by the invention is an end-to-end training and predicting method, is very simple, easy to use and efficient in the process of model training and attribute prediction, and is an advantage which is not possessed by the JRL method.
Finally, it should be noted that the above-mentioned embodiments are only for illustrating the technical solutions of the present invention and not for limiting, although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered by the claims of the present invention.
Claims (7)
1. A construction method of a pedestrian attribute identification network based on a recurrent neural network attention model is characterized by comprising the following steps:
a first convolution neural network for extracting the pedestrian whole-body image characteristics N (x) by using the pedestrian original whole-body image as input;
using the pedestrian whole-body image feature N (x) as the first inputIn, attention heat map A of the last time focused attribute groupt-1(x) As a second input, an attention heat map A of the set of attributes of interest at the current time is outputt(x) And pedestrian feature H passing through local highlightt(x) A recurrent neural network of (a);
using partially highlighted pedestrian features Ht(x) Outputting a second convolutional neural network of the attribute prediction probability of the current attention group as input;
the partially highlighted pedestrian feature Ht(x) Attention heat map A using the set of attributes that were focused on at the previous timet-1(x) The calculation formula is obtained by acting on the pedestrian whole-body image characteristic N (x) as follows:
Ht(x)＝At(x) oN (x) + N (x), and t represents the t-th group.
2. The method of constructing a pedestrian attribute identification network of claim 1 wherein a batch regularization operation is used on the attribute prediction probability outputs.
3. The method for constructing a pedestrian attribute recognition network according to any one of claims 1 to 2, comprising: for each different attribute group of the same original pedestrian whole-body image, the state of a memory unit of the recurrent neural network is determined by the characteristics of the locally highlighted pedestrians of all the predicted attribute groups; sharing the weight value of the first convolution neural network at different prediction moments; the second convolutional neural network shares the weights for different predicted times.
4. The method of claim 3, wherein the pedestrian attribute recognition network is trained using a weighted Sigmoid cross entropy loss function, wherein the loss function is as follows:
Wj＝exp(pj)
in the above formula, pjRepresenting the proportion of the positive examples number of the attribute j in the training set, w representing the learning weight of the positive examples,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe label is the label of the jth attribute of the ith sample, N is the total number of training samples, and K is the total number of attributes to be identified.
5. A pedestrian attribute identification method based on a recurrent neural network attention model is characterized by comprising the following steps:
s1, acquiring a certain number of pedestrian images with attributes to be identified, marking whether the images have certain attributes or not, and acquiring a data set which can be used for training the identification effect of the attributes of the pedestrians; grouping all the labeled attributes according to semantic and spatial neighbor relations;
s2, constructing a pedestrian attribute identification network based on the recurrent neural network attention model according to any one of claims 1 to 4 by combining an increment network and a convolutional recurrent neural network;
s3, defining a loss function required by training the pedestrian attribute recognition network, and training the pedestrian attribute recognition network constructed in the step S2 by using the training data set obtained in the step S1;
and S4, identifying the attributes in the pedestrian image to be identified by using the pedestrian attribute identification network trained in the step S3.
6. The pedestrian attribute identification method according to claim 5, wherein the step S2 includes:
s2-1, extracting the original pedestrian whole-body image by using an inclusion network to obtain the pedestrian whole-body image characteristic N (x);
s2-2, at the moment i, calculating the relation of the current moment by using the characteristics N (x) of the whole-body image of the pedestrian and a convolution cyclic neural networkAttention heatmap A of property groups of notest(x) And storing the historical information in a memory unit of the convolution cyclic neural network;
s2-3, use attention heatmap At(x) Acting on the pedestrian whole-body image feature N (x) to obtain the partially highlighted pedestrian feature Ht(x) The calculation formula is as follows:
Ht(x)＝At(x) oN (x) + N (x); t represents the t-th group;
s2-4, using the locally highlighted feature Ht(x) And carrying out attribute identification on the t-th group of attributes and outputting the prediction probability of the group of attributes.
7. The pedestrian attribute identification method according to claim 5 or 6, wherein the loss function defined in the step S3 is as follows:
Wj＝exp(pj)
in the above formula, pjRepresenting the proportion of the positive examples number of the attribute j in the training set, w representing the learning weight of the positive examples,
probability, y, representing whether the model output model includes the jth attribute for the ith sample predictionijThe label is the label of the jth attribute of the ith sample, N is the total number of training samples, and K is the total number of attributes to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810616398.XA CN108921051B (en) | 2018-06-15 | 2018-06-15 | Pedestrian attribute identification network and technology based on cyclic neural network attention model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810616398.XA CN108921051B (en) | 2018-06-15 | 2018-06-15 | Pedestrian attribute identification network and technology based on cyclic neural network attention model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108921051A CN108921051A (en) | 2018-11-30 |
| CN108921051B true CN108921051B (en) | 2022-05-20 |
Family
ID=64421633
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810616398.XA Active CN108921051B (en) | 2018-06-15 | 2018-06-15 | Pedestrian attribute identification network and technology based on cyclic neural network attention model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921051B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711386B (en) * | 2019-01-10 | 2020-10-09 | 北京达佳互联信息技术有限公司 | Method and device for obtaining recognition model, electronic equipment and storage medium |
| CN109815902B (en) * | 2019-01-24 | 2021-04-27 | 北京邮电大学 | Method, device and equipment for acquiring pedestrian attribute region information |
| CN109886154A (en) * | 2019-01-30 | 2019-06-14 | 电子科技大学 | Most pedestrian's appearance attribute recognition methods according to collection joint training based on Inception V3 |
| CN109886241A (en) * | 2019-03-05 | 2019-06-14 | 天津工业大学 | Driver fatigue detection based on shot and long term memory network |
| CN110032952B (en) * | 2019-03-26 | 2020-11-10 | 西安交通大学 | Road boundary point detection method based on deep learning |
| CN110110601B (en) * | 2019-04-04 | 2023-04-25 | 深圳久凌软件技术有限公司 | Video pedestrian re-recognition method and device based on multi-time space attention model |
| CN109978077B (en) * | 2019-04-08 | 2021-03-12 | 南京旷云科技有限公司 | Visual recognition method, device and system and storage medium |
| CN110163296B (en) * | 2019-05-29 | 2020-12-18 | 北京达佳互联信息技术有限公司 | Image recognition method, device, equipment and storage medium |
| CN110287836B (en) * | 2019-06-14 | 2021-10-15 | 北京迈格威科技有限公司 | Image classification method and device, computer equipment and storage medium |
| CN110458215B (en) * | 2019-07-30 | 2023-03-24 | 天津大学 | Pedestrian attribute identification method based on multi-temporal attention model |
| CN110688888B (en) * | 2019-08-02 | 2022-08-05 | 杭州未名信科科技有限公司 | Pedestrian attribute identification method and system based on deep learning |
| CN110569779B (en) * | 2019-08-28 | 2022-10-04 | 西北工业大学 | Pedestrian attribute identification method based on pedestrian local and overall attribute joint learning |
| CN110633421B (en) * | 2019-09-09 | 2020-08-11 | 北京瑞莱智慧科技有限公司 | Feature extraction, recommendation, and prediction methods, devices, media, and apparatuses |
| CN110598631B (en) * | 2019-09-12 | 2021-04-02 | 合肥工业大学 | Pedestrian attribute identification method and system based on sequence context learning |
| CN110705474B (en) * | 2019-09-30 | 2022-05-03 | 清华大学 | Pedestrian attribute identification method and device |
| CN111539341B (en) * | 2020-04-26 | 2023-09-22 | 香港中文大学（深圳） | Target positioning method, device, electronic equipment and medium |
| CN113706437B (en) * | 2020-05-21 | 2024-03-15 | 国网智能科技股份有限公司 | Method and system for diagnosing defects of fine-granularity bolts of power transmission line |
| CN112580494A (en) * | 2020-12-16 | 2021-03-30 | 北京影谱科技股份有限公司 | Method and device for identifying and tracking personnel in monitoring video based on deep learning |
| CN114694177B (en) * | 2022-03-10 | 2023-04-28 | 电子科技大学 | Fine-grained character attribute identification method based on multi-scale feature and attribute association mining |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104050685A (en) * | 2014-06-10 | 2014-09-17 | 西安理工大学 | Moving target detection method based on particle filtering visual attention model |
| CN106971154A (en) * | 2017-03-16 | 2017-07-21 | 天津大学 | Pedestrian's attribute forecast method based on length memory-type recurrent neural network |
| CN107341462A (en) * | 2017-06-28 | 2017-11-10 | 电子科技大学 | A kind of video classification methods based on notice mechanism |
| 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 |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016065534A1 (en) * | 2014-10-28 | 2016-05-06 | 中国科学院自动化研究所 | Deep learning-based gait recognition method |
| US9830529B2 (en) * | 2016-04-26 | 2017-11-28 | Xerox Corporation | End-to-end saliency mapping via probability distribution prediction |
-
2018
- 2018-06-15 CN CN201810616398.XA patent/CN108921051B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104050685A (en) * | 2014-06-10 | 2014-09-17 | 西安理工大学 | Moving target detection method based on particle filtering visual attention model |
| CN106971154A (en) * | 2017-03-16 | 2017-07-21 | 天津大学 | Pedestrian's attribute forecast method based on length memory-type recurrent neural network |
| CN107341462A (en) * | 2017-06-28 | 2017-11-10 | 电子科技大学 | A kind of video classification methods based on notice mechanism |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108921051A (en) | 2018-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921051B (en) | Pedestrian attribute identification network and technology based on cyclic neural network attention model | |
| CN107704877B (en) | Image privacy perception method based on deep learning | |
| CN111259786B (en) | Pedestrian re-identification method based on synchronous enhancement of appearance and motion information of video | |
| CN103971386B (en) | A kind of foreground detection method under dynamic background scene | |
| CN111723693B (en) | Crowd counting method based on small sample learning | |
| CN105389562B (en) | A kind of double optimization method of the monitor video pedestrian weight recognition result of space-time restriction | |
| CN107633226B (en) | Human body motion tracking feature processing method | |
| CN110222592B (en) | Construction method of time sequence behavior detection network model based on complementary time sequence behavior proposal generation | |
| JP6397379B2 (en) | CHANGE AREA DETECTION DEVICE, METHOD, AND PROGRAM | |
| CN109886141A (en) | A kind of pedestrian based on uncertainty optimization discrimination method again | |
| CN111339818B (en) | Face multi-attribute recognition system | |
| CN113536972B (en) | Self-supervision cross-domain crowd counting method based on target domain pseudo label | |
| CN110298297A (en) | Flame identification method and device | |
| CN111178284A (en) | Pedestrian re-identification method and system based on spatio-temporal union model of map data | |
| CN116342894B (en) | GIS infrared feature recognition system and method based on improved YOLOv5 | |
| CN113297936A (en) | Volleyball group behavior identification method based on local graph convolution network | |
| CN113283368B (en) | Model training method, face attribute analysis method, device and medium | |
| Kadim et al. | Deep-learning based single object tracker for night surveillance. | |
| Lin et al. | Optimal CNN-based semantic segmentation model of cutting slope images | |
| Li et al. | Fire flame image detection based on transfer learning | |
| CN108154199B (en) | High-precision rapid single-class target detection method based on deep learning | |
| Pillai et al. | Fine-Tuned EfficientNetB4 Transfer Learning Model for Weather Classification | |
| CN115705706A (en) | Video processing method, video processing device, computer equipment and storage medium | |
| Delahoz et al. | A deep-learning-based floor detection system for the visually impaired | |
| CN115661539A (en) | Less-sample image identification method embedded with uncertainty information |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181130 Assignee: CSIC PRIDE(Nanjing)Intelligent Equipment System Co.,Ltd Assignor: TSINGHUA University Contract record no.: X2023320000119 Denomination of invention: Pedestrian attribute recognition network and technology based on recurrent neural network attention model Granted publication date: 20220520 License type: Common License Record date: 20230323 |