CN107301376A - A kind of pedestrian detection method stimulated based on deep learning multilayer - Google Patents
A kind of pedestrian detection method stimulated based on deep learning multilayer Download PDFInfo
- Publication number
- CN107301376A CN107301376A CN201710385952.3A CN201710385952A CN107301376A CN 107301376 A CN107301376 A CN 107301376A CN 201710385952 A CN201710385952 A CN 201710385952A CN 107301376 A CN107301376 A CN 107301376A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- msub
- mrow
- multilayer
- frame
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of pedestrian detection method stimulated based on deep learning multilayer, for after given video monitoring and the target that need to be detected, marking the position that target occurs in video.Specifically include following steps：Pedestrian's data set for training objective detection model is obtained, and defines algorithm target；Position deviation and apparent semanteme to pedestrian target are modeled；Modeling result in step S2, which sets up pedestrian's multilayer, stimulates network model；The pedestrian position in monitoring image is detected using the detection model.Pedestrian detection of the present invention suitable for real video monitoring image, has preferably effect and robustness in face of all kinds of complex situations.
Description
Technical field
The invention belongs to computer vision field, a kind of particularly pedestrian detection stimulated based on deep learning multilayer
Method.
Background technology
Since 20 end of the centurys, with the development of computer vision, intelligent video treatment technology is widely paid close attention to and ground
Study carefully.Pedestrian detection is the important and challenging task of one of which, and its target is accurately to detect in video monitoring image
The position of pedestrian.The problem has very high application value in the field such as video monitoring and intelligent robot, is largely senior regard
The basis of feel task.But same, the problem has larger challenge, and one is how to express target area information；How two be
The extraction of candidate region and target classification unified Modeling are optimized, these challenges are proposed to the performance and robustness of respective algorithms
Higher requirement.
General pedestrian detection algorithm is divided into three parts：1st, the candidate region that target is included in input picture is found out.2nd, base
In candidate region manual extraction target signature.3rd, Detection task is realized using sorting algorithm to feature.This kind of method is primarily present
Following problem：1) it is based on traditional visual signature, and these visual signatures can only express the visual information of lower level, but row
People's Detection task needs model to possess the semantic understanding ability of higher level of abstraction；2) extraction of candidate region and the classification of feature do not have
End-to-end study optimization；3) combination is not stimulated by multilayer based on the feature that deep learning is extracted, target signature is not abstract enough
It is abundant.
The content of the invention
To solve the above problems, it is an object of the invention to provide a kind of pedestrian detection stimulated based on deep learning multilayer
Method, for detecting the pedestrian position in given monitoring image.This method is based on deep neural network, the depth stimulated using multilayer
Spend visual signature and characterize target area information, pedestrian detection is modeled using Faster R-CNN frameworks, can better adapt to true
Complex situations in real video monitoring scene.
To achieve the above object, the technical scheme is that：
A kind of pedestrian detection method stimulated based on deep learning multilayer, is comprised the following steps：
S1, pedestrian's data set for training objective detection model is obtained, and define algorithm target；
S2, the position deviation to pedestrian target and apparent semanteme are modeled；
S3, the modeling result in step S2, which set up pedestrian's multilayer, stimulates network model；
S4, use the detection model detection monitoring image in pedestrian position.
Further, in step S1, described pedestrian's data set for training objective detection model, including pedestrian image
Xtrain, the pedestrian position B manually marked；
Defining algorithm target is：Detect the pedestrian position P in a width monitoring image X.
Further, in step S2, position deviation and apparent semanteme to pedestrian target are modeled and specifically included：
S21, according to pedestrian's data set XtrainPosition deviation is modeled with pedestrian position P：
Wherein, x, y are the middle point coordinates of pedestrian's box label, and w, h is the width and length of pedestrian's box label, xa,yaIt is pedestrian
The coordinate of candidate frame, wa,haIt is the width and length of pedestrian candidate frame；txFor pedestrian's frame x coordinate relative to callout box x coordinate
Deviation correspondence mark width of frame ratio, tyFor deviation correspondence callout box of the y-coordinate relative to callout box y-coordinate of pedestrian's frame
The ratio of length, twFor ratio of the width relative to mark width of frame of pedestrian's frame, thFor pedestrian's frame length relative to callout box
The ratio of length；
S22, according to pedestrian's data set XtrainApparent semanteme is modeled with pedestrian position P：
S=<w,d>
Wherein s represents projection values of the feature d on projection vector w, and w is pedestrian's weight projection vector, and d is that pedestrian's feature is retouched
State son,<.,.>It is that inner product operation is accorded with, and p (C=k | d) it is softmax functions, represent to belong to the probable value of kth class；sjIt is characterized d
Projection value on j-th of projection vector w；C is the discrete random variable that value number is k；J is whole projection vector w jth
Individual w index.
Further, in step S3, the modeling result in step S2, which sets up pedestrian's multilayer, stimulates network model specific
Including：
S31, set up multilayer and stimulate convolutional neural networks, the input of neutral net is that a width monitoring image X and pedestrian mark
Frame B, is output as the probable value p of correspondence pedestrian candidate frame, and the pedestrian position deviation O in X；The representation of neutral net is
Map X → (p, O)；
S32, sub- mapping X → p use soft maximum Softmax loss functions, are expressed as
Lcls(X,Y；θ)=- ∑jYjLogp (C | d) formula (3)
Wherein Y is binary set, if belonging to kth class, respective value is 1, and remaining is 0；Lcls(X,Y；θ) represent whole instruction
Practice the softmax loss functions of data set；
S33, sub- mapping X → O use Euclid's loss function, are expressed as
Lloc(t, v)=∑ismooth(ti,vi)
Wherein tiIt is pedestrian position deviation label, viIt is pedestrian position deflection forecast value；I represents i-th of training sample；
S34, the loss function of whole multilayer stimulation neutral net are
L=Lcls+LlocFormula (5)
Whole neutral net is trained under loss function L using stochastic gradient descent and back-propagation algorithm.
Further, in step S4, the pedestrian position in detection monitoring image includes：Monitoring image X to be detected is defeated
Enter the neutral net trained, the candidate frame probable value according to its output determines whether pedestrian, finally according to the position of prediction
Deviation O corrections obtain pedestrian position P.
The present invention is applied to the pedestrian detection method of video monitoring scene, compared to existing pedestrian detection method, has
Following beneficial effect：
First, pedestrian detection method of the invention sets up model based on depth convolutional neural networks.The present invention is by candidate regions
The generation in domain and the classification of feature, which are unified in same network frame, learns optimization, improves the final effect of method.
Secondly, multilayer proposed by the present invention stimulate algorithm can more feature-rich abstracting power, while the Algorithm Learning
The feature gone out is so that grader learns the classifying rules of more robust.
The present invention is applied to the pedestrian detection method of video monitoring scene, has well in intelligent video analysis system
Application value, can effectively improve efficiency and the degree of accuracy of pedestrian detection.For example, in traffic video monitoring, row of the invention
People's detection method can quickly and correctly detect all pedestrian positions, and pedestrian's search mission for after provides data, greatly
Release human resources.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the pedestrian detection method applied to video monitoring scene of the present invention；
Fig. 2 stimulates the loss function schematic diagram of neutral net for the whole multilayer of the present invention.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
On the contrary, the present invention covers any replacement done in the spirit and scope of the present invention being defined by the claims, repaiied
Change, equivalent method and scheme.Further, in order that the public has a better understanding to the present invention, below to the thin of the present invention
It is detailed to describe some specific detail sections in section description.Part without these details for a person skilled in the art
Description can also understand the present invention completely.
With reference to Fig. 1, in the preferred embodiment, a kind of pedestrian detection side stimulated based on deep learning multilayer
Method, comprises the following steps：
First, pedestrian's data set for training objective detection model, including pedestrian image X are obtainedtrain, manually mark
Pedestrian position B；
Defining algorithm target is：Detect the pedestrian position P in a width monitoring image X.
Secondly, position deviation and apparent semanteme to pedestrian target are modeled and specifically included：
The first step, according to pedestrian's data set XtrainPosition deviation is modeled with pedestrian position P：
Wherein, x, y are the middle point coordinates of pedestrian's box label, and w, h is the width and length of pedestrian's box label, xa,yaIt is pedestrian
The coordinate of candidate frame, wa,haIt is the width and length of pedestrian candidate frame；txFor pedestrian's frame x coordinate relative to callout box x coordinate
Deviation correspondence mark width of frame ratio, tyFor deviation correspondence callout box of the y-coordinate relative to callout box y-coordinate of pedestrian's frame
The ratio of length, twFor ratio of the width relative to mark width of frame of pedestrian's frame, thFor pedestrian's frame length relative to callout box
The ratio of length；
Second step, according to pedestrian's data set XtrainApparent semanteme is modeled with pedestrian position P：
S=<w,d>
Wherein s represents projection values of the feature d on projection vector w, and w is pedestrian's weight projection vector, and d is that pedestrian's feature is retouched
State son,<.,.>It is that inner product operation is accorded with, and p (C=k | d) it is softmax functions, represent to belong to the probable value of kth class；sjIt is characterized d
Projection value on j-th of projection vector w；C is the discrete random variable that value number is k；J is whole projection vector w jth
Individual w index.
Afterwards, according to the detection model of appeal modeling result pre-training billboard target.Specifically include：
The first step, setting up multilayer stimulates convolutional neural networks, and the input of neutral net is that a width monitoring image X and pedestrian mark
Frame B is noted, the probable value p of correspondence pedestrian candidate frame, and the pedestrian position deviation O in X is output as；So as to the knot of neutral net
Structure can be expressed as mapping X → (p, O)；
Second step, sub- mapping X → p is expressed as using soft maximum (Softmax) loss function
Lcls(X,Y；θ)=- ∑jYjLogp (C | d) formula (3)
Wherein Y is binary set, if belonging to kth class, respective value is 1, and remaining is 0；Lcls(X,Y；θ) represent whole instruction
Practice the softmax loss functions of data set；
3rd step, sub- mapping X → O uses Euclid's loss function, is expressed as
Lloc(t, v)=∑ismooth(ti,vi)
Wherein tiIt is pedestrian position deviation label, viIt is pedestrian position deflection forecast value, i represents i-th of training sample.
4th step, with reference to Fig. 2, the loss function of whole multilayer stimulation neutral net is
L=Lcls+LlocFormula (5)
Whole neutral net is trained under loss function L using stochastic gradient descent and back-propagation algorithm.
Finally, the pedestrian in monitoring image is detected using the detection model trained.Specifically include：Will pretreatment
Good image, which is put into multilayer, stimulates calculating in detection framework.Multilayer stimulates detection framework to extract candidate frame with 3 RPN networks,
The characteristic information that each RPN networks are utilized is different, and candidate frame size and yardstick are also different obtained from.First obtain each
The candidate frame of RPN network extractions, 300 candidate regions are filtrated to get according to respective confidence level size.Then by 3 RPN networks
In candidate region merge, obtain 900 candidate regions.Arrange, be filtrated to get most from big to small then according to classification confidence
300 whole object candidate areas.Whether the candidate frame class probability value according to its output, which is more than given threshold value, is filtered candidate frame,
Overlapping detection block is eliminated using non-maxima suppression algorithm simultaneously, is corrected finally according to the position deviation O of prediction
To pedestrian position P.
In above-described embodiment, the position deviation and apparent semanteme of pedestrian detection method of the invention first to pedestrian target are entered
Row modeling.On this basis, former problem is converted into multi-task learning problem, and pedestrian detection is set up based on deep neural network
Model.Finally, the pedestrian position in monitoring image is detected using the detection model trained.
By above technical scheme, the embodiment of the present invention has been developed a kind of many based on deep learning based on depth learning technology
The pedestrian detection algorithm that layer is stimulated.The present invention can effectively model the position deviation and apparent semantic information of target simultaneously, so that
Detect accurate pedestrian position.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
Any modifications, equivalent substitutions and improvements made within refreshing and principle etc., should be included in the scope of the protection.
Claims (5)
1. a kind of pedestrian detection method stimulated based on deep learning multilayer, it is characterised in that comprise the following steps：
S1, pedestrian's data set for training objective detection model is obtained, and define algorithm target；
S2, the position deviation to pedestrian target and apparent semanteme are modeled；
S3, the modeling result in step S2, which set up pedestrian's multilayer, stimulates network model；
S4, use the detection model detection monitoring image in pedestrian position.
2. the pedestrian detection method as claimed in claim 1 stimulated based on deep learning multilayer, it is characterised in that step S1
In, described pedestrian's data set for training objective detection model, including pedestrian image Xtrain, the pedestrian position manually marked
B；
Defining algorithm target is：Detect the pedestrian position P in a width monitoring image X.
3. the pedestrian detection method as claimed in claim 2 stimulated based on deep learning multilayer, it is characterised in that step S2
In, position deviation and apparent semanteme to pedestrian target are modeled and specifically included：
S21, according to pedestrian's data set XtrainPosition deviation is modeled with pedestrian position P：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>t</mi>
<mi>x</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>a</mi>
</msub>
</mrow>
<msub>
<mi>w</mi>
<mi>a</mi>
</msub>
</mfrac>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>t</mi>
<mi>y</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>a</mi>
</msub>
</mrow>
<msub>
<mi>h</mi>
<mi>a</mi>
</msub>
</mfrac>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, x, y are the middle point coordinates of pedestrian's box label, and w, h is the width and length of pedestrian's box label, xa,yaIt is pedestrian candidate
The coordinate of frame, wa,haIt is the width and length of pedestrian candidate frame；txIt is inclined relative to callout box x coordinate for the x coordinate of pedestrian's frame
The ratio of difference correspondence mark width of frame, tyFor deviation correspondence callout box length of the y-coordinate relative to callout box y-coordinate of pedestrian's frame
Ratio, twFor ratio of the width relative to mark width of frame of pedestrian's frame, thFor pedestrian's frame length relative to callout box length
Ratio；
S22, according to pedestrian's data set XtrainApparent semanteme is modeled with pedestrian position P：
S=<w,d>
Wherein s represents projection values of the feature d on projection vector w, and w is pedestrian's weight projection vector, and d is the description of pedestrian's feature
Son,<.,.>It is that inner product operation is accorded with, and p (C=k | d) it is softmax functions, represent to belong to the probable value of kth class；sjD is characterized to exist
Projection value on j-th of projection vector w；C is the discrete random variable that value number is k；J is j-th of whole projection vector w
W index.
4. the pedestrian detection method as claimed in claim 3 stimulated based on deep learning multilayer, it is characterised in that step S3
In, the modeling result in step S2, which sets up pedestrian's multilayer, stimulates network model to specifically include：
S31, set up multilayer and stimulate convolutional neural networks, the input of neutral net is a width monitoring image X and pedestrian callout box B,
It is output as the probable value p of correspondence pedestrian candidate frame, and the pedestrian position deviation O in X；The representation of neutral net is mapping
X→(p,O)；
S32, sub- mapping X → p use soft maximum Softmax loss functions, are expressed as
<mrow>
<mi>p</mi>
<mrow>
<mo>(</mo>
<mi>C</mi>
<mo>=</mo>
<mi>k</mi>
<mo>|</mo>
<mi>d</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<msub>
<mi>s</mi>
<mi>k</mi>
</msub>
</msup>
<mo>/</mo>
<msub>
<mi>&Sigma;</mi>
<mi>j</mi>
</msub>
<msup>
<mi>e</mi>
<msub>
<mi>s</mi>
<mi>j</mi>
</msub>
</msup>
</mrow>
Lcls(X,Y；θ)=- ∑jYjLogp (C | d) formula (3)
Wherein Y is binary set, if belonging to kth class, respective value is 1, and remaining is 0；Lcls(X,Y；θ) represent whole training number
According to the softmax loss functions of collection；
S33, sub- mapping X → O use Euclid's loss function, are expressed as
Lloc(t, v)=∑ismooth(ti,vi)
Wherein tiIt is pedestrian position deviation label, viIt is pedestrian position deflection forecast value；I represents i-th of training sample；
S34, the loss function of whole multilayer stimulation neutral net are
L=Lcls+LlocFormula (5)
Whole neutral net is trained under loss function L using stochastic gradient descent and back-propagation algorithm.
5. the pedestrian detection method as claimed in claim 4 stimulated based on deep learning multilayer, it is characterised in that step S4
In, the pedestrian position in detection monitoring image includes：The neutral net that monitoring image X inputs to be detected are trained, foundation
Its candidate frame probable value exported determines whether pedestrian, and the position deviation O corrections finally according to prediction obtain pedestrian position P.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710385952.3A CN107301376B (en) | 2017-05-26 | 2017-05-26 | Pedestrian detection method based on deep learning multi-layer stimulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710385952.3A CN107301376B (en) | 2017-05-26 | 2017-05-26 | Pedestrian detection method based on deep learning multi-layer stimulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107301376A true CN107301376A (en) | 2017-10-27 |
| CN107301376B CN107301376B (en) | 2021-04-13 |
Family
ID=60138099
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710385952.3A Active CN107301376B (en) | 2017-05-26 | 2017-05-26 | Pedestrian detection method based on deep learning multi-layer stimulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107301376B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108446662A (en) * | 2018-04-02 | 2018-08-24 | 电子科技大学 | A kind of pedestrian detection method based on semantic segmentation information |
| CN108537117A (en) * | 2018-03-06 | 2018-09-14 | 哈尔滨思派科技有限公司 | A kind of occupant detection method and system based on deep learning |
| CN110163224A (en) * | 2018-01-23 | 2019-08-23 | 天津大学 | It is a kind of can on-line study auxiliary data mask method |
| CN110969657A (en) * | 2018-09-29 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Gun and ball coordinate association method and device, electronic equipment and storage medium |
| CN111178267A (en) * | 2019-12-30 | 2020-05-19 | 成都数之联科技有限公司 | Video behavior identification method for monitoring illegal fishing |
| CN111476089A (en) * | 2020-03-04 | 2020-07-31 | 上海交通大学 | Pedestrian detection method, system and terminal based on multi-mode information fusion in image |
| CN111523478A (en) * | 2020-04-24 | 2020-08-11 | 中山大学 | Pedestrian image detection method acting on target detection system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016149881A1 (en) * | 2015-03-20 | 2016-09-29 | Intel Corporation | Object recogntion based on boosting binary convolutional neural network features |
| CN106022237A (en) * | 2016-05-13 | 2016-10-12 | 电子科技大学 | Pedestrian detection method based on end-to-end convolutional neural network |
| CN106250812A (en) * | 2016-07-15 | 2016-12-21 | 汤平 | A kind of model recognizing method based on quick R CNN deep neural network |
| WO2017062610A1 (en) * | 2015-10-06 | 2017-04-13 | Evolv Technologies, Inc. | Augmented machine decision making |
-
2017
- 2017-05-26 CN CN201710385952.3A patent/CN107301376B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016149881A1 (en) * | 2015-03-20 | 2016-09-29 | Intel Corporation | Object recogntion based on boosting binary convolutional neural network features |
| WO2017062610A1 (en) * | 2015-10-06 | 2017-04-13 | Evolv Technologies, Inc. | Augmented machine decision making |
| CN106022237A (en) * | 2016-05-13 | 2016-10-12 | 电子科技大学 | Pedestrian detection method based on end-to-end convolutional neural network |
| CN106250812A (en) * | 2016-07-15 | 2016-12-21 | 汤平 | A kind of model recognizing method based on quick R CNN deep neural network |
Non-Patent Citations (5)
| Title |
|---|
| JIE LIU ET AL.: "Deep Convolutional Neural Networks for Pedestrian Detection with Skip Pooling", 《 2017 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS》 * |
| JIFENG DAI ET AL.: "R-FCN: Object Detection via Region-based Fully Convolutional Networks", 《ARXIV:1605.06409V2》 * |
| ROSS GIRSHICK: "Fast R-CNN", 《ARXIV:1504.08083V2》 * |
| ZHAOWEI CAI ET AL.: "A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection", 《EUROPEAN CONFERENCE ON COMPUTER VISION》 * |
| 任少卿: "基于特征共享的高效物体检测", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163224A (en) * | 2018-01-23 | 2019-08-23 | 天津大学 | It is a kind of can on-line study auxiliary data mask method |
| CN110163224B (en) * | 2018-01-23 | 2023-06-20 | 天津大学 | Auxiliary data labeling method capable of online learning |
| CN108537117A (en) * | 2018-03-06 | 2018-09-14 | 哈尔滨思派科技有限公司 | A kind of occupant detection method and system based on deep learning |
| CN108446662A (en) * | 2018-04-02 | 2018-08-24 | 电子科技大学 | A kind of pedestrian detection method based on semantic segmentation information |
| CN110969657A (en) * | 2018-09-29 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Gun and ball coordinate association method and device, electronic equipment and storage medium |
| CN110969657B (en) * | 2018-09-29 | 2023-11-03 | 杭州海康威视数字技术股份有限公司 | Gun ball coordinate association method and device, electronic equipment and storage medium |
| CN111178267A (en) * | 2019-12-30 | 2020-05-19 | 成都数之联科技有限公司 | Video behavior identification method for monitoring illegal fishing |
| CN111476089A (en) * | 2020-03-04 | 2020-07-31 | 上海交通大学 | Pedestrian detection method, system and terminal based on multi-mode information fusion in image |
| CN111476089B (en) * | 2020-03-04 | 2023-06-23 | 上海交通大学 | Pedestrian detection method, system and terminal for multi-mode information fusion in image |
| CN111523478A (en) * | 2020-04-24 | 2020-08-11 | 中山大学 | Pedestrian image detection method acting on target detection system |
| CN111523478B (en) * | 2020-04-24 | 2023-04-28 | 中山大学 | Pedestrian image detection method acting on target detection system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107301376B (en) | 2021-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107301376A (en) | A kind of pedestrian detection method stimulated based on deep learning multilayer | |
| CN109635694B (en) | Pedestrian detection method, device and equipment and computer readable storage medium | |
| CN106447658B (en) | Conspicuousness object detection method based on global and local convolutional network | |
| CN107862261A (en) | Image people counting method based on multiple dimensioned convolutional neural networks | |
| CN104182772B (en) | A kind of gesture identification method based on deep learning | |
| CN104143079B (en) | The method and system of face character identification | |
| CN107506722A (en) | One kind is based on depth sparse convolution neutral net face emotion identification method | |
| CN107123123A (en) | Image segmentation quality evaluating method based on convolutional neural networks | |
| CN106682697A (en) | End-to-end object detection method based on convolutional neural network | |
| CN107704877A (en) | A kind of image privacy cognitive method based on deep learning | |
| CN106778835A (en) | The airport target by using remote sensing image recognition methods of fusion scene information and depth characteristic | |
| CN110111340A (en) | The Weakly supervised example dividing method cut based on multichannel | |
| CN107341452A (en) | Human bodys' response method based on quaternary number space-time convolutional neural networks | |
| CN105005774A (en) | Face relative relation recognition method based on convolutional neural network and device thereof | |
| CN107633511A (en) | A kind of blower fan vision detection system based on own coding neutral net | |
| CN106529448A (en) | Method for performing multi-visual-angle face detection by means of integral channel features | |
| CN107134144A (en) | A kind of vehicle checking method for traffic monitoring | |
| CN107945153A (en) | A kind of road surface crack detection method based on deep learning | |
| CN106295506A (en) | A kind of age recognition methods based on integrated convolutional neural networks | |
| CN107506786A (en) | A kind of attributive classification recognition methods based on deep learning | |
| CN105678278A (en) | Scene recognition method based on single-hidden-layer neural network | |
| CN108280397A (en) | Human body image hair detection method based on depth convolutional neural networks | |
| CN107766890A (en) | The improved method that identification segment learns in a kind of fine granularity identification | |
| CN104240256A (en) | Image salient detecting method based on layering sparse modeling | |
| CN108256462A (en) | A kind of demographic method in market monitor video |
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 |