CN107729805A - The neutral net identified again for pedestrian and the pedestrian based on deep learning recognizer again - Google Patents
The neutral net identified again for pedestrian and the pedestrian based on deep learning recognizer again Download PDFInfo
- Publication number
- CN107729805A CN107729805A CN201710780179.0A CN201710780179A CN107729805A CN 107729805 A CN107729805 A CN 107729805A CN 201710780179 A CN201710780179 A CN 201710780179A CN 107729805 A CN107729805 A CN 107729805A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- neural network
- identification
- image
- cnn
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 8
- 230000007935 neutral effect Effects 0.000 title abstract 3
- 230000009466 transformation Effects 0.000 claims abstract description 24
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 22
- 238000013528 artificial neural network Methods 0.000 claims description 74
- 238000013527 convolutional neural network Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 2
- 244000287680 Garcinia dulcis Species 0.000 abstract 1
- 210000005036 nerve Anatomy 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 15
- 238000000605 extraction Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000000034 method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 2
- 241000271897 Viperidae Species 0.000 description 1
- 210000000746 body region Anatomy 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
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
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/14—Transformations for image registration, e.g. adjusting or mapping for alignment of images
- G06T3/147—Transformations for image registration, e.g. adjusting or mapping for alignment of images using affine transformations
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (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)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the neutral net identified again for pedestrian and the recognizer again of the pedestrian based on deep learning.The neutral net includes：Inputted using the original whole body images of pedestrian as first and export the first nerves network of the first identification feature；Affine transformation image using the human body image extracted from the original whole body images of pedestrian inputs as second and exports the nervus opticus network of the second identification feature, wherein, human body comprises at least head, trunk and four limbs, and the first identification feature and the second identification feature are combined as total identification feature.Pedestrian's characteristic matching ability with more robust, so as to improve correct recognition rata and/or reduce misclassification rate.
Description
Technical Field
The present invention relates to the field of convolutional neural networks and image recognition. More particularly, the present invention relates to a neural network for pedestrian re-recognition and a pedestrian re-recognition algorithm based on deep learning.
Background
With the popularization of video monitoring technology, the role of pedestrian re-identification technology becomes increasingly important, and the pedestrian re-identification technology can help people to automatically complete the task of searching for specific personnel from massive image or video data.
Feature extraction and feature matching using convolutional neural networks are two important components of pedestrian re-identification technology. However, there are usually large posture changes and complicated viewing angle changes between different images of the pedestrian acquired by different cameras, which both greatly increase the difficulty of matching the pedestrian with the pedestrian re-recognition algorithm.
At least in order to solve the above technical problems, a new pedestrian re-identification algorithm capable of adapting to the posture change needs to be provided, so that the algorithm can better adapt to the posture change of the pedestrian, extract more robust features, and thus improve the final correct identification rate and/or reduce the false identification rate.
Disclosure of Invention
The purpose of the invention is realized by the following technical scheme.
A neural network for pedestrian re-identification, comprising:
a first neural network that uses an original whole-body image of a pedestrian as a first input and outputs a first recognition feature;
a second neural network that uses an affine transformation image of a human body part image extracted from an original whole-body image of a pedestrian as a second input and outputs a second recognition feature,
the human body part at least comprises a head, a trunk and four limbs, and the first identification feature and the second identification feature are combined into a total identification feature.
The neural network for pedestrian re-recognition according to the present invention further comprises:
a feature embedding sub-neural network (FEN) includes a Pose Transformation Network (PTN) for applying an affine transformation to each body part in a body part map to obtain a more robust affine transformation image of the body part map image.
The neural network for pedestrian re-recognition according to the present invention further comprises:
a feature weighting sub-neural network (FWN) for weighting and biasing the second identifying features output by the second neural network to combine the first identifying features and the second identifying features into a total identifying feature.
According to the neural network for pedestrian re-identification of the present invention, the first neural network includes a first Convolutional Neural Network (CNN) and a second CNN, and the second neural network includes a third CNN and a fourth CNN, wherein the first CNN and the third CNN share a weight.
According to the neural network for pedestrian re-recognition of the present invention, the second CNN and the fourth CNN adopt independent weights.
The neural network for pedestrian re-identification according to the invention is characterized in that the first neural network and the second neural network each further comprise a Convolution (CONV) layer and a Global Average Pooling (GAP) layer at the respective output.
The pedestrian re-identification algorithm based on deep learning comprises the following steps:
constructing a neural network for pedestrian re-identification according to the above;
training the neural network for pedestrian re-recognition using a pedestrian re-recognition dataset;
pedestrian re-recognition is performed using a trained neural network for pedestrian re-recognition.
The invention has the advantages that: the pedestrian feature matching method has more robust pedestrian feature matching capability, so that the correct recognition rate can be improved and/or the false recognition rate can be reduced.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the specific embodiments. The drawings are only for purposes of illustrating the particular embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
fig. 1 shows a conceptual schematic diagram of a first neural network for pedestrian re-identification according to an embodiment of the present invention.
Fig. 2 shows an overall schematic diagram of a second type of neural network for pedestrian re-identification according to an embodiment of the invention.
Fig. 3 shows a schematic block diagram of an FEN comprised by a second neural network for pedestrian re-identification according to an embodiment of the present invention.
Fig. 4 shows a partial processing result of the FEN included in the second neural network for pedestrian re-recognition according to the embodiment of the present invention.
Fig. 5 shows a schematic block diagram of a PTN in an FEN included in a second neural network for pedestrian re-recognition according to an embodiment of the present invention.
Fig. 6 shows a detailed view of a third neural network for pedestrian re-identification according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Fig. 1 shows a conceptual schematic diagram of a first neural network 100 for pedestrian re-identification according to an embodiment of the present invention.
As shown in fig. 1, a first type of neural network 100 for pedestrian re-identification includes a first neural network 101 and a second neural network 103.
As shown in fig. 1, the first neural network 101 uses an original whole-body image of a pedestrian as a first input and outputs a first recognition feature Fglobal。
The second neural network 103 uses an affine transformation image of the human body part image extracted from the original whole-body image of the pedestrian as a second input and outputs a second recognition feature Fpart。
Wherein the human body part at least comprises a head, a trunk and four limbs, and a first identification characteristic FglobalAnd a second identifying feature FpartCombined into a total recognition feature Ffusion。
For example, as known to those skilled in the art, the original whole-body image of the pedestrian may be selected from pedestrian re-recognition data sets such as CUHK 03, Market1501, and VIPeR for training the weight of the first neural network or testing the correct recognition rate.
For example, as known to those skilled in the art, the body part image may be manually extracted from the original whole-body image of the pedestrian by an artificial manner, and the affine transformation parameters are manually adjusted to obtain an affine transformation image, which is used as the second input for training the weight of the second neural network or testing the correct recognition rate.
The extraction of the body region image can also be performed by a pose estimation algorithm commonly used in the art (e.g., a pose estimation algorithm based on a Full Convolution Network (FCN)), and specific technical solutions refer to j.long, e.shell, and article "full volumetric networks for segmentation", published by t.darrell in 2015 on CVPR.
Fig. 2 shows an overall schematic block diagram of a second type of neural network 200 for pedestrian re-identification according to an embodiment of the present invention.
As shown in fig. 2, in order to realize a function of automatically acquiring an affine transformation image of a human body part image from an original whole body image of a pedestrian, a second neural network 200 for pedestrian re-recognition includes a FEN 201 in addition to a first neural network 101 and a second neural network 103.
The FEN 201 is configured to automatically acquire a human body part image from an original whole-body image of a pedestrian, then perform affine transformation on the human body part image using PTN, which takes the original whole-body image of the pedestrian as an input, and output the image of the human body part image after arithmetic processing, which has undergone PTN affine transformation, to the second neural network 103.
Fig. 3 shows a schematic block diagram of an FEN 201 comprised by a second neural network 200 for pedestrian re-identification according to an embodiment of the present invention.
As shown in fig. 3, FEN 201 includes an FCN-based pose estimation module 301 using existing techniques (e.g., as disclosed by j. long et al and which may be separately trained and incorporated into FEN 201), a human body position image extraction module 303 using techniques customary in the art, and first, second, third, fourth, and fifth PTNs 305, 307, 309, 311, 313. The first PTN 305, the second PTN 307, the third PTN 309, the fourth PTN 311, and the fifth PTN 313 are respectively configured to perform affine transformation on the images of the left arm, the right arm, the torso, the left leg, and the right leg of the pedestrian extracted by the human body part image extraction module 303, and finally obtain affine transformation images of human body parts other than the head. Finally, the FEN 201 combines the head original image and the affine transformation image of each human body part other than the head, and outputs the combined image as a PTN affine-transformed image.
That is, the FEN 201 includes PTN for applying affine transformation to each human body part in the human body parts map to obtain an affine transformation image of a more robust human body parts map image.
Fig. 4 shows a partial processing result of the FEN 201 included in the second neural network 200 for pedestrian re-recognition according to the embodiment of the present invention.
As shown in fig. 4, its subgraph (a) is a pair of original whole-body images of the same pedestrian (different view angles are adopted for camera shooting). Sub-diagram (b) is a diagram of the human body posture estimation result obtained by the human body posture estimation module 301 using FCN, for example. Sub-diagram (c) is a pair of joint indication diagrams labeled 14 human joints, for example, acquired by pose estimation module 301 using FCN. Sub-diagram (d) is a pixel region corresponding to each of 6 human body parts including a head, a torso, and limbs, which are acquired by the human posture estimation module 301 using FCN, for example. Sub-diagram (e) is, for example, an image including only the above 6 human body parts acquired by using the human body part image extraction module 303. The sub-image (f) is, for example, an affine transformation image obtained by processing the image in the sub-image (e) using PTN. Optionally, the image in (e) may be rotated and scaled, and the image in (f) may be normalized.
Fig. 5 shows a schematic block diagram of a PTN in the FEN 201 included in the second neural network for pedestrian re-recognition according to the embodiment of the present invention.
As shown in fig. 5, the PTN network (i.e., the first PTN 305, the second PTN 307, the third PTN 309, the fourth PTN 311, and the fifth PTN 313 shown in fig. 3) is used to acquire parameters a for performing affine transformation of an image of limbs (left arm, right arm, left leg, and right leg) and an image of a trunk by trainingTheta-limbs or trunk. The PTN network was modified based on the Spatial Transform Network (STN) disclosed in the article Spatial transform networks 2015 published by m.jaderberg, k.simonyan, a.zisserman et al on NIPS.
As shown in formula (1), the affine transformation parameter A for the human body part image obtained by the PTN networkθIs 6-dimensional, of which theta1、θ2、θ4And theta5Is the scaling and rotation parameters, and theta3And theta6Is a translation parameter. (x)s，ys) Is the pixel coordinate in the original image of the body part, and (x)t，yt) Is the pixel coordinates in the affine transformation image of the human body part.
Optionally, as shown in fig. 2, the second neural network 200 for pedestrian re-identification further comprises FWN 203.
FWN 203 is used to weight and bias the second identifying features output by the second neural network 103 to combine the first identifying features and the second identifying features into a total identifying feature. The weighting and biasing are implemented by equation (2).
tanh(Fpart⊙W+B) (2)
Wherein,w and B are weight and bias vector, respectively, which are associated with the second identifying feature FpartIs the same dimension, ⊙ denotes a Hadamard product operation.
Also, to implement error back-propagation, the following gradient update formula may be defined for FWN 203 for training FWN 203.
Wherein f isi∈Ffusion(i＝1，2…m+n)，gj∈|Fglobal(j＝1，2…m)，pk∈Fpart(k＝1，2…n)，wkI ∈ W (k ═ 1, 2 … n), B ∈ B (k ═ 1, 2 … n), m and n are F, respectivelyglobalAnd FpartDimension (d) of (a).
Fig. 6 shows a detailed schematic diagram of a third neural network 300 for pedestrian re-identification according to an embodiment of the present invention.
As shown in FIG. 6, the first neural network used by the third neural network 300 includes a first CNN and a second CNN (CNN in FIG. 6)g) The second neural network includes a third CNN and a fourth CNN (CNN in FIG. 6)p) And the first CNN and the third CNN share the weight value.
Alternatively, as shown in fig. 6, the second CNN and the fourth CNN used by the third neural network 300 use independent weights.
Optionally, as shown in fig. 6, the first neural network and the second neural network used by the third neural network 300 each further include a CONV layer and a GAP layer at respective output ends to map output features to appropriate dimensions for subsequent processing and may support input images of different sizes.
The invention also provides a pedestrian re-identification algorithm based on deep learning and combined with the neural network for pedestrian re-identification. The pedestrian re-identification algorithm based on deep learning comprises the following steps:
a neural network for pedestrian re-identification according to the above is constructed.
Training the neural network for pedestrian re-recognition using a pedestrian re-recognition dataset.
Pedestrian re-recognition is performed using a trained neural network for pedestrian re-recognition.
The technical scheme of the invention has better capability of feature extraction and feature matching, thereby improving the final correct recognition rate and/or reducing the false recognition rate.
The above description is only an exemplary embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (7)
1. A neural network for pedestrian re-identification, comprising:
a first neural network that uses an original whole-body image of a pedestrian as a first input and outputs a first recognition feature;
a second neural network that uses an affine transformation image of a human body part image extracted from an original whole-body image of a pedestrian as a second input and outputs a second recognition feature,
the human body part at least comprises a head, a trunk and four limbs, and the first identification feature and the second identification feature are combined into a total identification feature.
2. The neural network for pedestrian re-identification according to claim 1, further comprising:
the characteristic embedding sub-neural network FEN comprises a posture transformation neural network PTN, and the PTN is used for applying affine transformation to each human body part in the human body part map to obtain a more robust affine transformation image of the human body part map image.
3. The neural network for pedestrian re-identification according to claim 1 or 2, further comprising:
the feature weighting sub-neural network FWN is used to weight and bias the second identifying features output by the second neural network, thereby combining the first identifying features and the second identifying features into a total identifying feature.
4. The neural network for pedestrian re-identification according to claim 3, wherein the first neural network comprises a first Convolutional Neural Network (CNN) and a second CNN, and the second neural network comprises a third CNN and a fourth CNN, wherein the first CNN and the third CNN share weight values.
5. The neural network for pedestrian re-identification, as claimed in claim 4, wherein the second CNN and the fourth CNN use independent weights.
6. The neural network for pedestrian re-identification according to claim 1 or 2, characterized in that the first and second neural networks each further comprise a convolutional CONV layer and a global average pooling GAP layer at the respective output.
7. A pedestrian re-identification algorithm based on deep learning, comprising:
constructing a neural network for pedestrian re-identification according to any one of claims 1-6;
training the neural network for pedestrian re-recognition using a pedestrian re-recognition dataset;
pedestrian re-recognition is performed using a trained neural network for pedestrian re-recognition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710780179.0A CN107729805B (en) | 2017-09-01 | 2017-09-01 | The neural network identified again for pedestrian and the pedestrian based on deep learning recognizer again |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710780179.0A CN107729805B (en) | 2017-09-01 | 2017-09-01 | The neural network identified again for pedestrian and the pedestrian based on deep learning recognizer again |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107729805A true CN107729805A (en) | 2018-02-23 |
| CN107729805B CN107729805B (en) | 2019-09-13 |
Family
ID=61205446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710780179.0A Active CN107729805B (en) | 2017-09-01 | 2017-09-01 | The neural network identified again for pedestrian and the pedestrian based on deep learning recognizer again |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107729805B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109101901A (en) * | 2018-07-23 | 2018-12-28 | 北京旷视科技有限公司 | Human action identification and its neural network generation method, device and electronic equipment |
| CN109886113A (en) * | 2019-01-17 | 2019-06-14 | 桂林远望智能通信科技有限公司 | A kind of spacious view pedestrian recognition methods again based on region candidate network |
| CN109934081A (en) * | 2018-08-29 | 2019-06-25 | 厦门安胜网络科技有限公司 | A kind of pedestrian's attribute recognition approach, device and storage medium based on deep neural network |
| CN109977837A (en) * | 2019-03-20 | 2019-07-05 | 常熟理工学院 | Pedestrian based on human body attitude invariant features recognition methods again |
| CN110543817A (en) * | 2019-07-25 | 2019-12-06 | 北京大学 | Pedestrian re-identification method based on posture guidance feature learning |
| CN110688873A (en) * | 2018-07-04 | 2020-01-14 | 上海智臻智能网络科技股份有限公司 | Multi-target tracking method and face recognition method |
| CN110956576A (en) * | 2018-09-27 | 2020-04-03 | 北京小米移动软件有限公司 | Image processing method, device, equipment and storage medium |
| US10783393B2 (en) | 2017-06-20 | 2020-09-22 | Nvidia Corporation | Semi-supervised learning for landmark localization |
| CN111920436A (en) * | 2020-07-08 | 2020-11-13 | 浙江大学 | Dual-tracer PET (positron emission tomography) separation method based on multi-task learning three-dimensional convolutional coding and decoding network |
| CN112990144A (en) * | 2021-04-30 | 2021-06-18 | 德鲁动力科技(成都)有限公司 | Data enhancement method and system for pedestrian re-identification |
| CN113255615A (en) * | 2021-07-06 | 2021-08-13 | 南京视察者智能科技有限公司 | Pedestrian retrieval method and device for self-supervision learning |
| TWI740624B (en) * | 2019-10-28 | 2021-09-21 | 中國商深圳市商湯科技有限公司 | Image processing method, device and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140177946A1 (en) * | 2012-12-21 | 2014-06-26 | Electronics and Telecommunicatidons Research Institute | Human detection apparatus and method |
| CN104915643A (en) * | 2015-05-26 | 2015-09-16 | 中山大学 | Deep-learning-based pedestrian re-identification method |
| CN106022392A (en) * | 2016-06-02 | 2016-10-12 | 华南理工大学 | Deep neural network sample automatic accepting and rejecting training method |
| CN106529442A (en) * | 2016-10-26 | 2017-03-22 | 清华大学 | Pedestrian identification method and apparatus |
| CN106845415A (en) * | 2017-01-23 | 2017-06-13 | 中国石油大学(华东) | A kind of pedestrian based on deep learning becomes more meticulous recognition methods and device |
| CN106951872A (en) * | 2017-03-24 | 2017-07-14 | 江苏大学 | A kind of recognition methods again of the pedestrian based on unsupervised depth model and hierarchy attributes |
-
2017
- 2017-09-01 CN CN201710780179.0A patent/CN107729805B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140177946A1 (en) * | 2012-12-21 | 2014-06-26 | Electronics and Telecommunicatidons Research Institute | Human detection apparatus and method |
| CN104915643A (en) * | 2015-05-26 | 2015-09-16 | 中山大学 | Deep-learning-based pedestrian re-identification method |
| CN106022392A (en) * | 2016-06-02 | 2016-10-12 | 华南理工大学 | Deep neural network sample automatic accepting and rejecting training method |
| CN106529442A (en) * | 2016-10-26 | 2017-03-22 | 清华大学 | Pedestrian identification method and apparatus |
| CN106845415A (en) * | 2017-01-23 | 2017-06-13 | 中国石油大学(华东) | A kind of pedestrian based on deep learning becomes more meticulous recognition methods and device |
| CN106951872A (en) * | 2017-03-24 | 2017-07-14 | 江苏大学 | A kind of recognition methods again of the pedestrian based on unsupervised depth model and hierarchy attributes |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10783394B2 (en) | 2017-06-20 | 2020-09-22 | Nvidia Corporation | Equivariant landmark transformation for landmark localization |
| US10783393B2 (en) | 2017-06-20 | 2020-09-22 | Nvidia Corporation | Semi-supervised learning for landmark localization |
| CN110688873A (en) * | 2018-07-04 | 2020-01-14 | 上海智臻智能网络科技股份有限公司 | Multi-target tracking method and face recognition method |
| CN109101901A (en) * | 2018-07-23 | 2018-12-28 | 北京旷视科技有限公司 | Human action identification and its neural network generation method, device and electronic equipment |
| CN109101901B (en) * | 2018-07-23 | 2020-10-27 | 北京旷视科技有限公司 | Human body action recognition method and device, neural network generation method and device and electronic equipment |
| CN109934081A (en) * | 2018-08-29 | 2019-06-25 | 厦门安胜网络科技有限公司 | A kind of pedestrian's attribute recognition approach, device and storage medium based on deep neural network |
| CN110956576A (en) * | 2018-09-27 | 2020-04-03 | 北京小米移动软件有限公司 | Image processing method, device, equipment and storage medium |
| CN110956576B (en) * | 2018-09-27 | 2024-03-22 | 北京小米移动软件有限公司 | Image processing method, device, equipment and storage medium |
| CN109886113A (en) * | 2019-01-17 | 2019-06-14 | 桂林远望智能通信科技有限公司 | A kind of spacious view pedestrian recognition methods again based on region candidate network |
| CN109977837A (en) * | 2019-03-20 | 2019-07-05 | 常熟理工学院 | Pedestrian based on human body attitude invariant features recognition methods again |
| CN110543817A (en) * | 2019-07-25 | 2019-12-06 | 北京大学 | Pedestrian re-identification method based on posture guidance feature learning |
| TWI740624B (en) * | 2019-10-28 | 2021-09-21 | 中國商深圳市商湯科技有限公司 | Image processing method, device and storage medium |
| CN111920436A (en) * | 2020-07-08 | 2020-11-13 | 浙江大学 | Dual-tracer PET (positron emission tomography) separation method based on multi-task learning three-dimensional convolutional coding and decoding network |
| CN112990144A (en) * | 2021-04-30 | 2021-06-18 | 德鲁动力科技(成都)有限公司 | Data enhancement method and system for pedestrian re-identification |
| CN113255615A (en) * | 2021-07-06 | 2021-08-13 | 南京视察者智能科技有限公司 | Pedestrian retrieval method and device for self-supervision learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107729805B (en) | 2019-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107729805B (en) | The neural network identified again for pedestrian and the pedestrian based on deep learning recognizer again | |
| CN109684924B (en) | Face living body detection method and device | |
| Wang et al. | 360sd-net: 360 stereo depth estimation with learnable cost volume | |
| CN111862296B (en) | Three-dimensional reconstruction method, three-dimensional reconstruction device, three-dimensional reconstruction system, model training method and storage medium | |
| GB2608975A (en) | Person identification across multiple captured images | |
| CN103607554A (en) | Fully-automatic face seamless synthesis-based video synthesis method | |
| CN107767339B (en) | Binocular stereo image splicing method | |
| CN112528902B (en) | Video monitoring dynamic face recognition method and device based on 3D face model | |
| CN111260543A (en) | Underwater image splicing method based on multi-scale image fusion and SIFT features | |
| CN108171735B (en) | Billion pixel video alignment method and system based on deep learning | |
| Wang et al. | A segmentation based robust deep learning framework for multimodal retinal image registration | |
| CN110263605A (en) | Pedestrian's dress ornament color identification method and device based on two-dimension human body guise estimation | |
| Charco et al. | Deep learning based camera pose estimation in multi-view environment | |
| CN111553939A (en) | Image registration algorithm of multi-view camera | |
| CN111898571A (en) | Action recognition system and method | |
| Ruan et al. | Image stitching algorithm based on SURF and wavelet transform | |
| Zheng et al. | Online subspace learning from gradient orientations for robust image alignment | |
| CN114708617A (en) | Pedestrian re-identification method and device and electronic equipment | |
| CN111582036A (en) | Cross-view-angle person identification method based on shape and posture under wearable device | |
| CN111126250A (en) | Pedestrian re-identification method and device based on PTGAN | |
| Mo et al. | A Robust Infrared and Visible Image Registration Method for Dual Sensor UAV System | |
| CN110321452A (en) | A kind of image search method based on direction selection mechanism | |
| CN111047513B (en) | Robust image alignment method and device for cylindrical panorama stitching | |
| CN114066954A (en) | Feature extraction and registration method for multi-modal images | |
| Alzohairy et al. | Image mosaicing based on neural networks |
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 |