CN109376693A - Method for detecting human face and system - Google Patents
Method for detecting human face and system Download PDFInfo
- Publication number
- CN109376693A CN109376693A CN201811398883.0A CN201811398883A CN109376693A CN 109376693 A CN109376693 A CN 109376693A CN 201811398883 A CN201811398883 A CN 201811398883A CN 109376693 A CN109376693 A CN 109376693A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- frame
- module
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to technical field of computer vision, discloses a kind of method for detecting human face and system, the detection accuracy to solve traditional Face datection algorithm are not able to satisfy the requirement of application.The present invention first passes through acquisition and includes the image of face, then demarcates to the human face region in image；The face frame coordinate of calibration is converted again；It reuses clustering algorithm and chooses anchor frame；Re-optimization deep learning Face datection network；Again based on the deep learning Face datection network after Face Sample Storehouse and optimization, training face detection model；Image to be detected is inputted again, to feature is extracted after image to be detected pretreatment, based on trained faceform, is returned face frame using the anchor frame that cluster obtains, is detected face, final output testing result simultaneously depicts face frame position in the picture.The present invention is suitable for Face datection.
Description
Technical field
The present invention relates to technical field of computer vision, in particular to method for detecting human face and system.
Background technique
With the continuous development of computer vision technique, human face detection tech tends to the mature and rapid marketization, as people
The committed step of face recognizer is widely applied in fields such as gate inhibition, attendance, testimony of a witness unification verifying, security protections.So-called people
Face detection, exactly gives any one picture, finds wherein with the presence or absence of one or more faces, and returns in picture everyone
The location and range of face.For Face datection according to the difference of application scenarios, can be divided into has constraint and without two kinds of situations of constraint.There is constraint
Situation refers to that face number is single, background is simple, deflection angle is small, it is unobstructed under the conditions of Face datection, basic face attendance
It is this kind of to belong to constrained situation with gate inhibition etc., performance requirement is sufficient for traditional innovatory algorithm；It is without constraint Face datection
Refer to, face dimensional variation is big, and quantity is more, and posture multiplicity is blocked, and illumination effect, expression shape change is big, and makeup camouflage etc. is related to
It is monitored in real time to security protection, intelligent video human face analysis, the scenes such as crowd's quantity statistics not only want the precision of Face datection algorithm
The real-time asked very high, and to ensure to run, the detection accuracy of traditional Face datection algorithm are not able to satisfy the requirement of application.
With the rise of deep learning algorithm, human face detection tech is increased to new level.Deep learning method and biography
Algorithm of uniting is different, disobeys selected characteristic by hand, transmits autonomous selection face characteristic by neural network.Deep neural network mould
Type is made of input layer, hidden layer (multilayer), output layer, and web results are deeper, and parameter is more, causes Face datection model larger,
And need to run in powerful computing platform (usually gpu), it is difficult to realize based on the real-time inspection under embedded cpu platform
It surveys.With the raising of the operational capability of the further investigation and embedded platform of light efficient depth network, so that being based on depth
The Face datection algorithm of study runs at embedded cpu and is possibly realized.
Summary of the invention
The technical problem to be solved by the present invention is a kind of method for detecting human face and system are provided, to solve traditional people
The detection accuracy of face detection algorithm is not able to satisfy the requirement of application.
To solve the above problems, the technical solution adopted by the present invention is that:
Method for detecting human face includes the following steps:
S001: acquisition includes the image of face, is demarcated to the human face region in image；
S002: the face frame coordinate of calibration is converted；
S003: anchor frame is chosen using clustering algorithm；
S004: optimization deep learning Face datection network；
S005: based on the deep learning Face datection network after Face Sample Storehouse and step S004 optimization, training face inspection
Survey model；
S006: input image to be detected, to feature is extracted after image to be detected pretreatment, based on step S005 training
Faceform returns face frame using the anchor frame that step S003 is clustered, detects face.
Further, the content of step S001 calibration includes: the top left co-ordinate and face square of human face region boundary rectangle
The length and width of shape frame.
Further, the face frame coordinate of calibration is converted in step S002, is specifically included:
S002_1: according to step S001 demarcate content calculate face frame centre coordinate (x_center, y_center) with
And the long h_rect of face frame and width w_rect；
S002_2: to step S002_1 calculate numerical value be normalized, obtain for training input x, y, w and
H, wherein x is face frame abscissa, and x=x_center/w_image, y are face frame ordinate, y=y_center/h_
Image, w are that face frame is long, and w=w_rect/image_width, h are face frame height, h=h_rect/h_image, w_image
It is respectively the width and height of image with h_image.
Further, step S003 chooses anchor frame using KMeans clustering algorithm, specifically includes:
Using KMeans clustering algorithm, the face rectangle frame marked is clustered out to the length and width of 6 kinds of suitable face frame ratios
Than.
Further, step S004 optimizes deep learning Face datection network, specifically includes:
S004_1: by the 3x3 convolution operation in former network, first using the convolution operation of 1x1, drop is logical in proportion as needed
Road reuses 3x3 convolution for channel and is upgraded to original quantity；
S004_2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level.
Face detection system, which is characterized in that including two parts of training and detection；Wherein, the training department point includes:
Sample process module, anchor frame choose module and training module；The detection part includes: image input module, image preprocessing
Module, face detection module and image output module.
Acquisition, calibration and face frame coordinate conversion of the sample process module for sample；
The anchor frame chooses module and is used to cluster out the face frame for being suitble to human face ratio；
The training module is used for design optimization depth Topological expansion deep learning Face datection network, and according to excellent
Deep learning Face datection network training human-face detector after change；
Described image input module is for obtaining image information；
For described image preprocessing module for zooming in and out to image, length and width are 32 multiple and identical；
The face detection module is for detecting the face for including in input picture；
Described image output module is for output test result and depicts face frame position in the picture.
Further, the specific steps of the sample process module progress face frame coordinate conversion include:
S002_1: the centre coordinate x_center and y_ that content calculates face frame are demarcated according to the sample process module
The long h_rect of center and face frame and width w_rect；
S002_2: being normalized step S002_1 evaluation, obtains for training input x, y, w and h,
In, x is face frame abscissa, and x=x_center/w_image, y are face frame ordinate, y=y_center/h_image, w
Long, the w=w_rect/image_width for face frame, h are face frame height, h=h_rect/h_image, w_image and h_
Image is respectively the width and height of image.
Further, the training module design optimization depth network structure, specifically includes:
S004_1: it first uses the convolution operation of 1x1 to drop in proportion as needed the 3x3 convolution operation in former network and leads to
Road reuses 3x3 convolution for channel and is upgraded to original quantity；
S004_2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level.
Further, the anchor frame, which chooses module and KMeans clustering algorithm can be used and cluster out, is suitble to the six of human face ratio
Class face frame.
The beneficial effects of the present invention are: the present invention uses deep learning model, dimension is risen after carrying out first dimensionality reduction to volume base
Operation reduces operand, compresses Face datection model size, improves operational efficiency.It is chosen simultaneously using KMeans clustering algorithm
It is most suitable for the anchor frame of face frame ratio out, improves detection accuracy.The system can be used under embedded cpu environment, can expire simultaneously
Computing resource is low and high two conditions of detection accuracy for foot consumption.
Detailed description of the invention
Fig. 1 is 1 method for detecting human face flow chart of embodiment
Fig. 2 is that 1 convolution of embodiment optimizes schematic diagram
Fig. 3 is 2 face detection system structure chart of embodiment.
Specific embodiment
Method for detecting human face and system of the invention are clearly and completely described below in conjunction with embodiment.
Embodiment 1
Embodiment 1 provides a kind of fast face detecting method, as shown in Figure 1, specific implementation step is as follows:
S001: sample collection calibration；
The S001 refers to that acquisition includes the image of face, demarcates to the human face region in image, usually face
The top left co-ordinate of region boundary rectangle and the length and width of face rectangle frame；
S002: the face frame coordinate of calibration is converted, comprising:
S002_1: according to step S001 demarcate content calculate face frame centre coordinate (x_center, y_center) with
And the long h_rect of face frame and width w_rect；
S002_2: to step S002_1 calculate numerical value be normalized, obtain for training input x, y, w and
H, wherein x is face frame abscissa, and x=x_center/w_image, y are face frame ordinate, y=y_center/h_
Image, w are that face frame is long, and w=w_rect/image_width, h are face frame height, h=h_rect/h_image, w_image
It is respectively the width and height of image with h_image；
S003: anchor frame is chosen using KMeans clustering algorithm.
Using KMeans clustering algorithm, the face rectangle frame marked is clustered out to the length and width of 6 kinds of suitable face frame ratios
Than；
S004: optimization deep learning Face datection network, comprising:
S004_1: by the 3x3 convolution operation in former network, first using the convolution operation of 1x1, drop is logical in proportion as needed
Road reuses 3x3 convolution for channel and is upgraded to original quantity, as shown in Fig. 2, the dimension of input feature map is 256 dimensions,
Seeking output dimension is also 256 dimensions, as shown in original convolution operation: the input of 256 dimensions is directly over one 3 × 3 × 256 convolution
Layer exports the feature map of one 256 dimension, then parameter amount is 256 × 3 × 3 × 256=589824；Convolution after such as improving
Shown in operation: the input of 256 dimensions first passes through one 1 × 1 × 64 convolutional layer, (no using one 3 × 3 × 64 convolutional layer
It is only limitted to be reduced to 64 dimensions, can adjusts as needed), finally pass through one 1 × 1 × 256 convolutional layer, 256 dimension of output, parameter
Amount are as follows: parameter amount and calculating can be greatly lowered in 256 × 1 × 1 × 64+64 × 3 × 3 × 64+64 × 1 × 1 × 256=69632
Amount；
S004_2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level；
S005: based on the deep learning Face datection network after Face Sample Storehouse and step S004 optimization, training face inspection
Survey model；
S006: input image to be detected, to feature is extracted after image to be detected pretreatment, based on step S005 training
Faceform returns face frame using the anchor frame that cluster obtains, detects face.In this step, anchor frame here refers to S003 step
Middle cluster obtains 6 kinds of anchor frames.
Embodiment 2
Embodiment 2 provides a kind of fast face detection system, as shown in figure 3, include two parts of training and detection,
Middle training department point includes: sample process module, anchor frame selection module and training module；Detection part include: image input module,
Image pre-processing module, face detection module and image output module.
Sample process module is used for the acquisition of sample, calibration and the conversion of face frame coordinate.
Anchor frame chooses module and is used to cluster out the face frame for being suitble to human face ratio；Anchor frame, which chooses module, can be used KMeans
Clustering algorithm clusters out the six class face frames for being suitble to human face ratio.
Training module is used to optimize deep learning Face datection network, and according to the deep learning Face datection net after optimization
Network training human-face detector；Optimize the method for deep learning Face datection network can include:
B1: first using the convolution operation of 1x1 to drop channel in proportion as needed the 3x3 convolution operation in former network, then
Channel is upgraded to original quantity using 3x3 convolution；
B2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level.
Image input module can be single image, video or image sequence for obtaining image information.
For image pre-processing module for zooming in and out to image, length and width are 32 multiple and identical.
Face detection module is for detecting the face for including in input picture.
Image output module is for output test result and depicts face frame position in the picture.
Based on the system of embodiment 2, embodiment 2 additionally provides a kind of method of fast face detection, the specific steps are as follows:
1, the acquisition of sample process module includes the image of face, demarcates, is demarcated to the human face region in image
The top left co-ordinate of human face region boundary rectangle and the length and width of face rectangle frame can be demarcated；
2, sample process module converts the face frame coordinate of calibration；
3, anchor frame is chosen module and is clustered out the six class anchor frames for being suitble to human face ratio using KMeans clustering algorithm；
4, training module optimizes deep learning Face datection network, and based on Face Sample Storehouse training face detection model；
5, image input module obtains image information, can be single image, video or image sequence, and pass through image
Preprocessing module zooms in and out image；
6, face detection module extracts characteristics of image, and the faceform based on the training of step 4 training module, uses step
The anchor frame that 3 clusters obtain returns face frame, face is detected, finally by image output module output test result and in the picture
Depict face frame position.
Claims (9)
1. method for detecting human face, which comprises the steps of:
S001: acquisition includes the image of face, is demarcated to the human face region in image；
S002: the face frame coordinate of calibration is converted；
S003: anchor frame is chosen using clustering algorithm；
S004: optimization deep learning Face datection network；
S005: based on the deep learning Face datection network after Face Sample Storehouse and step S004 optimization, training Face datection mould
Type；
S006: input image to be detected, to extraction feature after image to be detected pretreatment, the face based on step S005 training
Model returns face frame using the anchor frame that step S003 is clustered, detects face.
2. method for detecting human face as described in claim 1, which is characterized in that the content of step S001 calibration includes: face area
The top left co-ordinate of domain boundary rectangle and the length and width of face rectangle frame.
3. method for detecting human face as claimed in claim 2, which is characterized in that in step S002 to the face frame coordinate of calibration into
Row conversion, specifically includes:
S002_1: centre coordinate (x_center, y_center) and the people that content calculates face frame are demarcated according to step S001
The long h_rect and width w_rect of face frame；
S002_2: being normalized the step S002_1 numerical value calculated, obtain x, y, w and h for training input,
In, x is face frame abscissa, and x=x_center/w_image, y are face frame ordinate, y=y_center/h_image, w
Long, the w=w_rect/image_width for face frame, h are face frame height, h=h_rect/h_image, w_image and h_
Image is respectively the width and height of image.
4. method for detecting human face as described in claim 1, which is characterized in that step S003 is chosen using KMeans clustering algorithm
Anchor frame, specifically includes:
Using KMeans clustering algorithm, the face rectangle frame marked is clustered out to the length-width ratio of 6 kinds of suitable face frame ratios.
5. method for detecting human face as described in claim 1, which is characterized in that step S004 optimizes deep learning Face datection net
Network specifically includes:
S004_1: first using the convolution operation of 1x1 to drop channel in proportion as needed the 3x3 convolution operation in former network, then
Channel is upgraded to original quantity using 3x3 convolution；
S004_2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level.
6. face detection system, which is characterized in that including two parts of training and detection；Wherein, the training department point includes: sample
Processing module, anchor frame choose module and training module；The detection part includes: image input module, image preprocessing mould
Block, face detection module and image output module；
Acquisition, calibration and face frame coordinate conversion of the sample process module for sample；
The anchor frame chooses module and is used to cluster out the face frame for being suitble to human face ratio；
The training module is used to optimize deep learning Face datection network, and according to the deep learning Face datection net after optimization
Network training human-face detector；
Described image input module is for obtaining image information；
For described image preprocessing module for zooming in and out to image, length and width are 32 multiple and identical；
The face detection module is for detecting the face for including in input picture；
Described image output module is for output test result and depicts face frame position in the picture.
7. face detection system as claimed in claim 6, which is characterized in that the sample process module carries out face frame coordinate
The specific steps of conversion include:
A1: demarcating the centre coordinate x_center and y_center that content calculates face frame according to the sample process module, with
And the long h_rect of face frame and width w_rect；
A2: being normalized step S002_1 evaluation, obtains for training input x, y, w and h, wherein x is people
Face frame abscissa, x=x_center/w_image, y are face frame ordinate, and y=y_center/h_image, w are face frame
Long, w=w_rect/image_width, h are face frame height, and h=h_rect/h_image, w_image and h_image are respectively
The width and height of image.
8. face detection system as claimed in claim 6, which is characterized in that the training module design optimization depth network knot
Structure specifically includes:
B1: it first uses the convolution operation of 1x1 to drop channel in proportion as needed the 3x3 convolution operation in former network, reuses
Channel is upgraded to original quantity by 3x3 convolution；
B2: the route layer link structure in adjustment legacy network, it is corresponding with improved convolution sample level.
9. face detection system as claimed in claim 6, which is characterized in that the anchor frame is chosen module and clustered using KMeans
Algorithm clusters out the six class face frames for being suitble to human face ratio.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811398883.0A CN109376693A (en) | 2018-11-22 | 2018-11-22 | Method for detecting human face and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811398883.0A CN109376693A (en) | 2018-11-22 | 2018-11-22 | Method for detecting human face and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109376693A true CN109376693A (en) | 2019-02-22 |
Family
ID=65383370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811398883.0A Pending CN109376693A (en) | 2018-11-22 | 2018-11-22 | Method for detecting human face and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109376693A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110097114A (en) * | 2019-04-26 | 2019-08-06 | 新华三技术有限公司 | A kind of priori frame applied to neural network determines method and device |
| CN110210329A (en) * | 2019-05-13 | 2019-09-06 | 高新兴科技集团股份有限公司 | A kind of method for detecting human face, device and equipment |
| CN110287763A (en) * | 2019-04-11 | 2019-09-27 | 杭州电子科技大学 | A kind of candidate frame ratio optimization method towards ship seakeeping application |
| CN112017384A (en) * | 2020-08-05 | 2020-12-01 | 山东大学 | Automatic alarm method and system for real-time area monitoring |
| CN112084992A (en) * | 2020-09-18 | 2020-12-15 | 北京中电兴发科技有限公司 | Face frame selection method in face key point detection module |
| CN112150692A (en) * | 2020-10-14 | 2020-12-29 | 吴喜庆 | Access control method and system based on artificial intelligence |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100870724B1 (en) * | 2007-05-25 | 2008-11-27 | 인하대학교 산학협력단 | System for detecting image using t-test and method therefor |
| CN103390151A (en) * | 2012-05-08 | 2013-11-13 | 展讯通信（上海）有限公司 | Face detection method and device |
| CN106874868A (en) * | 2017-02-14 | 2017-06-20 | 北京飞搜科技有限公司 | A kind of method for detecting human face and system based on three-level convolutional neural networks |
| CN107818302A (en) * | 2017-10-20 | 2018-03-20 | 中国科学院光电技术研究所 | Non-rigid multi-scale object detection method based on convolutional neural network |
| CN108038474A (en) * | 2017-12-28 | 2018-05-15 | 深圳云天励飞技术有限公司 | Method for detecting human face, the training method of convolutional neural networks parameter, device and medium |
| CN108446694A (en) * | 2017-02-16 | 2018-08-24 | 杭州海康威视数字技术股份有限公司 | A kind of object detection method and device |
-
2018
- 2018-11-22 CN CN201811398883.0A patent/CN109376693A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100870724B1 (en) * | 2007-05-25 | 2008-11-27 | 인하대학교 산학협력단 | System for detecting image using t-test and method therefor |
| CN103390151A (en) * | 2012-05-08 | 2013-11-13 | 展讯通信（上海）有限公司 | Face detection method and device |
| CN106874868A (en) * | 2017-02-14 | 2017-06-20 | 北京飞搜科技有限公司 | A kind of method for detecting human face and system based on three-level convolutional neural networks |
| CN108446694A (en) * | 2017-02-16 | 2018-08-24 | 杭州海康威视数字技术股份有限公司 | A kind of object detection method and device |
| CN107818302A (en) * | 2017-10-20 | 2018-03-20 | 中国科学院光电技术研究所 | Non-rigid multi-scale object detection method based on convolutional neural network |
| CN108038474A (en) * | 2017-12-28 | 2018-05-15 | 深圳云天励飞技术有限公司 | Method for detecting human face, the training method of convolutional neural networks parameter, device and medium |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287763A (en) * | 2019-04-11 | 2019-09-27 | 杭州电子科技大学 | A kind of candidate frame ratio optimization method towards ship seakeeping application |
| CN110097114A (en) * | 2019-04-26 | 2019-08-06 | 新华三技术有限公司 | A kind of priori frame applied to neural network determines method and device |
| CN110097114B (en) * | 2019-04-26 | 2021-06-29 | 新华三技术有限公司 | Priori frame determination method and device applied to neural network |
| CN110210329A (en) * | 2019-05-13 | 2019-09-06 | 高新兴科技集团股份有限公司 | A kind of method for detecting human face, device and equipment |
| CN112017384A (en) * | 2020-08-05 | 2020-12-01 | 山东大学 | Automatic alarm method and system for real-time area monitoring |
| CN112084992A (en) * | 2020-09-18 | 2020-12-15 | 北京中电兴发科技有限公司 | Face frame selection method in face key point detection module |
| CN112084992B (en) * | 2020-09-18 | 2021-04-13 | 北京中电兴发科技有限公司 | Face frame selection method in face key point detection module |
| CN112150692A (en) * | 2020-10-14 | 2020-12-29 | 吴喜庆 | Access control method and system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109376693A (en) | Method for detecting human face and system | |
| CN110929578B (en) | Anti-shielding pedestrian detection method based on attention mechanism | |
| CN109635694B (en) | Pedestrian detection method, device and equipment and computer readable storage medium | |
| CN111460968B (en) | Unmanned aerial vehicle identification and tracking method and device based on video | |
| CN112633231B (en) | Fire disaster identification method and device | |
| CN114220035A (en) | Rapid pest detection method based on improved YOLO V4 | |
| CN110956094A (en) | RGB-D multi-mode fusion personnel detection method based on asymmetric double-current network | |
| RU2509355C2 (en) | Apparatus and method of classifying movement of objects in monitoring zone | |
| CN109284670A (en) | A kind of pedestrian detection method and device based on multiple dimensioned attention mechanism | |
| CN110348376A (en) | A kind of pedestrian's real-time detection method neural network based | |
| CN108960404B (en) | Image-based crowd counting method and device | |
| CN110119726A (en) | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model | |
| Li et al. | Sign language recognition based on computer vision | |
| CN105678231A (en) | Pedestrian image detection method based on sparse coding and neural network | |
| CN109543632A (en) | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features | |
| CN109508675A (en) | A kind of pedestrian detection method for complex scene | |
| CN110532961B (en) | Semantic traffic light detection method based on multi-scale attention mechanism network model | |
| TWI441096B (en) | Motion detection method for comples scenes | |
| CN110648311B (en) | Acne image focus segmentation and counting network model based on multitask learning | |
| CN110490252A (en) | A kind of occupancy detection method and system based on deep learning | |
| CN106023224A (en) | PCNN automatic segmentation method for microscopic image of traditional Chinese medicine | |
| CN109670405A (en) | A kind of complex background pedestrian detection method based on deep learning | |
| CN110263731B (en) | Single step human face detection system | |
| CN110287798A (en) | Vector network pedestrian detection method based on characteristic module and context fusion | |
| CN102693427A (en) | Method and device for forming detector for detecting images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190222 |