CN113705341A - Small-scale face detection method based on generation countermeasure network - Google Patents
Small-scale face detection method based on generation countermeasure network Download PDFInfo
- Publication number
- CN113705341A CN113705341A CN202110806601.1A CN202110806601A CN113705341A CN 113705341 A CN113705341 A CN 113705341A CN 202110806601 A CN202110806601 A CN 202110806601A CN 113705341 A CN113705341 A CN 113705341A
- Authority
- CN
- China
- Prior art keywords
- network
- face
- small
- scale
- generation
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 41
- 208000029152 Small face Diseases 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims description 16
- 230000004927 fusion Effects 0.000 claims description 9
- 230000014509 gene expression Effects 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 230000003042 antagnostic effect Effects 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000008485 antagonism Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a small-scale face detection method based on a generation countermeasure network, which relates to the technical field of face detection and solves the technical problem of poor precision of the existing small-scale face detection mode, and comprises the following steps: s1, inputting a noise image and a small face image in real collected data into a first generation countermeasure network to generate enough small face context area samples; s2, training a face detection network by using the small face context area samples and the small face images; s3, acquiring a workplace image, and acquiring a possibly existing small-scale face candidate area by utilizing a potential face area network; and S4, after super-resolution reconstruction of the small-scale face candidate region, inputting the reconstructed small-scale face candidate region into a face detection network to obtain a small-face position regression result. The invention generates enough samples of the context area of the small face by generating the confrontation network, solves the problem that the quantity of the existing small face samples cannot meet the requirement of network training, and further effectively improves the detection precision of the small-scale face by the super-resolution reconstruction network.
Description
Technical Field
The invention relates to the technical field of face detection, in particular to a small-scale face detection method based on a generation countermeasure network.
Background
With the progress of the times and the development of science and technology, the oil exploitation technology is greatly improved, and due to the particularity of the industry, the safety problem of the oil exploitation technology needs to be paid attention. The safety production management in oil exploitation mainly aims to ensure the safety of constructors and reduce casualties and property loss as much as possible. The geographic position of the oil and gas pump station is remote in reality, simple safety protection measures around and the personnel cannot be monitored all day by day, which is just a serious challenge to the monitoring problem of the safety protection personnel in current petroleum exploitation. Under the background of rapid development of the computer vision field, the face recognition technology is widely applied to the fields of identity verification, entrance guard recognition, monitoring and security protection and the like. The safety protection problem and the unable all-day monitoring problem of personnel that combine oil development, it is effective technological approach to carry out safety control to oil development personnel under unmanned aerial vehicle visual angle.
Under the oil exploitation environment, the problems of small face area, low resolution, incomplete details and the like exist in a face detection task under the view angle of an unmanned aerial vehicle, and the method is not suitable for directly inputting the current existing method model based on large face area, high resolution and complete details. In addition, in the adjustment and output network training stages in the cascade type face detection network, the number of face samples generated based on the area generation network cannot meet the data requirement of network training of the extremely small face, and the detection precision is poor.
Disclosure of Invention
The present invention is directed to solve the foregoing problems in the prior art, and an object of the present invention is to provide a small-scale face detection method based on a generative countermeasure network, which can improve detection accuracy.
The technical scheme of the invention is as follows: a small-scale face detection method based on generation of a confrontation network is characterized by comprising the following steps:
s1, inputting a noise image and a small face image in real collected data into a first generation countermeasure network to generate enough small face context area samples;
s2, training a face detection network by using the small face context area samples and the small face images;
s3, acquiring a workplace image, and acquiring a possibly existing small-scale face candidate area by utilizing a potential face area network;
and S4, inputting the small-scale face candidate area into the face detection network for classification and regression to obtain a small face position regression result.
As a further improvement, if the small-scale face candidate region in step S4 cannot be resolved to obtain a small face position regression result, the following steps are performed:
s5, inputting the small-scale face candidate region into a second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S6, inputting the reconstructed face candidate region into the face detection network for classification and regression to obtain a small face position regression result.
Further, the step S1 includes the following steps:
s1-1, inputting the noise image into a generating network G1 network of the first generating countermeasure network to obtain a sample of the face context area;
s1-2, inputting the mixed set of the sample of the face context area and the face image into the discrimination network D1 network of the first generation antagonizing network to obtain the discrimination of the sample of the face context area and the face image, and optimizing the parameters of the generation network G1 network and the discrimination network D1 network under the antagonism loss adjustment of the generation network G1 network and the discrimination network D1 network.
Further, the step S1-1 includes the following steps:
s1-1-1, the noise image satisfies z-N (0,1) and the pixel size is 48x48, a series of different outputs are obtained through split operation;
s1-1-2, adding the output of split operation to different ResBlock;
and S1-1-3, performing convolution of 1x1 before the last ResBlock to perform feature fusion and channel number reduction, and extracting feature expressions generated by the G1 network to obtain a sample of the face context area.
Further, the step S1-2 includes the following steps:
s1-2-1, increasing the number of channels by convolution of 1x1 on the mixed set of the sample of the lower region of the small face and the image of the small face so as to increase the feature expression capacity;
s1-2-2, performing feature extraction through ResBlock;
s1-2-3, performing information fusion by using convolution of 1x1 and performing discrimination of the sample of the face context area and the face image in the last full connection layer.
Further, the step S3 includes the following steps:
s3-1, detecting face candidate areas under different scales by using Resnet-50 as a feature extraction network;
s3-2, performing multilayer information fusion, connecting the high-level feature map with the shallow feature map after upsampling by using the characteristic that high-level semantics guide shallow feature information, and then uniformly outputting the size by using a full connection layer;
and S3-3, establishing a small-scale face candidate region which is possibly existed by carrying out regression on the face frame based on an anchor point mechanism in the specific feature layer.
Further, the step S5 includes the following steps:
s5-1, inputting the small-scale face candidate region into the generation network G2 network of the second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S5-2, judging whether the reconstructed face candidate region is generated data or not and whether the reconstructed face candidate region is a face sample or not by utilizing the judging network D2 network of the second generation countermeasure network.
Further, the step S5-1 includes the following steps:
s5-1-1, inputting the small-scale face candidate region into the first G2 network of the second generation countermeasure network to reconstruct to obtain a sample with higher resolution;
s5-1-2, inputting the sample with higher resolution into a second generation network G2 network of the second generation countermeasure network to reconstruct the human face candidate area from higher resolution to higher resolution;
s5-1-3, training the generating network G2 network.
Further, the step S5-2 includes the following steps:
s5-2-1, judging whether the reconstructed face candidate region is the classification loss of the generated data or not and whether the reconstructed face candidate region contains the classification loss of the face or not by the judging network D2 network of the second generation countermeasure network;
s5-2-2, training the discrimination network D2 network;
s5-2-3, the discrimination network D2 network jointly optimizes according to the classification loss of generated data, the classification loss of the human face and the confrontation loss of the generation network G2 network, so that the aim of reconstructing the small-scale human face candidate region is fulfilled.
Further, the noise image is an image that is randomly generated and normally distributed.
Advantageous effects
Compared with the prior art, the invention has the advantages that:
the method solves the problem that the small-scale face is difficult to process in the face detection task under the view angle of the unmanned aerial vehicle in the oil and gas exploitation environment by generating the countermeasure network, solves the problem that the small-scale face detection training sample is insufficient, carries out high-resolution reconstruction on the original target small-scale face, carries out refined expression on the small-scale face candidate region, reduces the difficulty of subsequent processing, and effectively improves the detection precision of the small-scale face, so that more small-scale faces can be detected, the false detection rate of the small-scale face is reduced, and effective guarantee is provided for the safety monitoring of personnel in the oil and gas exploitation environment.
Drawings
FIG. 1 is a flow chart of the detection of the present invention;
FIG. 2 is a flow chart of super-resolution reconstruction for low resolution in the present invention;
FIG. 3 is a flow chart of the discrimination network of the present invention;
FIG. 4 is a block diagram of the structure of the generation network G1 network in the present invention;
fig. 5 is a block diagram of the structure of the generation network G1 network ResBlock in the present invention;
FIG. 6 is a block diagram of the discrimination network D1 according to the present invention;
fig. 7 is a block diagram of the structure of the discrimination network D1 for ResBlock in the network in the present invention;
FIG. 8 is a diagram illustrating the effect of generating training samples using a generative confrontation network in the present invention;
fig. 9 is a diagram illustrating the effect of high resolution reconstruction using a generative countermeasure network in the present invention.
Detailed Description
The invention will be further described with reference to specific embodiments shown in the drawings.
Referring to fig. 1-9, a small-scale face detection method based on generation of a confrontation network includes the following steps:
s1, inputting a noise image and a small face image in real collected data into a first generation countermeasure network to generate enough small face context area samples; the small face image in the real collected data is the face image in the real photo;
s2, training a Face detection network by using samples of the context area of the small Face and the small Face image to solve the problem of insufficient number of small Face samples, wherein the Face detection network is a cascade type Face detection network or a Face-RCNN network;
s3, acquiring a workplace image, and acquiring a possibly existing small-scale face candidate area by utilizing a potential face area network; the workplace image can be an image of an indoor or outdoor workplace, such as an image shot by an unmanned aerial vehicle, an unmanned vehicle or other remote photographing equipment, and the face in the image is generally small;
and S4, inputting the small-scale face candidate area into a face detection network for classification and regression to obtain a small face position regression result.
Further, if the small-scale face candidate region in step S4 cannot be resolved to obtain a small face position regression result, the following steps are performed:
s5, inputting the small-scale face candidate region into a second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S6, inputting the reconstructed face candidate area into a face detection network for classification and regression to obtain a small face position regression result. If the result of the regression of the position of the small face is not obtained in step S6, detection failure information is output.
The step S1 includes the following steps:
s1-1, inputting the noise image into a generating network G1 network of a first generating countermeasure network to obtain a sample of the face context area;
s1-2, inputting the mixed set of the sample of the face context area and the face image into a discrimination network D1 network of a first generation antagonizing network to obtain discrimination of the sample of the face context area and the face image, and as shown in FIG. 3, optimizing parameters of a generation network G1 network and a discrimination network D1 network under the antagonistic loss adjustment of the generation network G1 network and the discrimination network D1 network.
The step S1-1 includes the following steps:
s1-1-1, the noise image satisfies z-N (0,1) and the pixel size is 48x48, a series of different outputs are obtained through split operation; the noise image is an image which is randomly generated and normally distributed, as shown in fig. 8;
s1-1-2, adding the output of split operation to different ResBlock;
s1-1-3, before the last ResBlock, performing convolution of 1x1 to perform feature fusion and channel number reduction, extracting feature expressions generated by the generated network G1 to obtain a sample of a face context area, and generating a network 1G network as shown in fig. 4, and a network G1 network ResBlock as shown in fig. 5.
The step S1-2 includes the following steps:
s1-2-1, increasing the number of channels by convolution of 1x1 on a mixed set of the sample of the lower region of the small face and the image of the small face so as to increase the feature expression capacity;
s1-2-2, performing feature extraction through ResBlock;
s1-2-3, performing information fusion by using convolution of 1x1, and performing discrimination between the sample of the lower region of the small face and the small face image at the last full connection layer, wherein the structural block diagram of the discrimination network D1 is shown in FIG. 6, and the structural block diagram of ResBlock in the discrimination network D1 is shown in FIG. 7.
The step S3 includes the following steps:
s3-1, detecting human face candidate areas under different scales by using Resnet-50 as a feature extraction network, and selecting features of three different layers Res3b, Res4b and Res5b in consideration of different scaling ratios of feature maps of the different layers to an original image;
s3-2, performing multilayer information fusion, connecting the high-level feature map with the shallow feature map after upsampling by using the characteristic that high-level semantics guide shallow feature information, and then uniformly outputting the size by using a full connection layer;
and S3-3, establishing a small-scale face candidate region which is possibly existed by carrying out regression on the face frame based on an anchor point mechanism in the specific feature layer.
The step S5 includes the following steps:
s5-1, inputting the small-scale face candidate region into a generation network G2 network of a second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S5-2, judging whether the reconstructed face candidate region is generated data or not and whether the reconstructed face candidate region is a face sample or not by utilizing a judgment network D2 of the second generation countermeasure network so as to achieve the purpose of judging the small-scale face candidate region which cannot be distinguished in the generation network.
The step S5-1 includes the following steps:
s5-1-1, inputting the small-scale face candidate region into a first generation network G2 network of a second generation countermeasure network to reconstruct to obtain a sample with higher resolution, wherein the sample with higher resolution is 2 times of the sample with original resolution;
s5-1-2, inputting the sample with higher resolution into a second generation network G2 network of a second generation countermeasure network to reconstruct the sample from higher resolution to obtain a reconstructed face candidate region; such as a double-resolution to quadruple-resolution reconstruction, i.e. reconstructing the face candidate region to 4 times samples at the original resolution, as shown in fig. 2.
S5-1-3, training to generate a network G2 network.
The step S5-2 includes the following steps:
s5-2-1, judging whether the reconstructed face candidate region is the classification loss of the generated data or not and whether the reconstructed face candidate region contains the classification loss of the face or not by the judgment network D2 network of the second generation confrontation network;
s5-2-2, training a discriminant network D2 network;
s5-2-3, the discrimination network D2 jointly optimizes according to the classification loss of the generated data, the classification loss of the human face and the confrontation loss of the generation network G2 network, and therefore the purpose of reconstructing the small-scale human face candidate region is achieved.
The invention solves the problem that the small-scale human face is difficult to process in the human face detection task under the view angle of the unmanned aerial vehicle in the oil and gas exploitation environment by means of generating the countermeasure network. The method solves the problem of small-scale face detection training, carries out high-resolution reconstruction on the original target small-scale face, and reduces the difficulty of subsequent processing, so that more extremely small-scale faces can be detected, and the false detection rate of the extremely small-scale faces is reduced. This provides effective guarantee to personnel's safety monitoring under the oil and gas exploitation environment.
In particular, by the above technical solution conceived herein, compared with the prior art, the following beneficial effects can be obtained:
(1) the method utilizes the generated confrontation network to generate the training samples, and effectively solves the problem that the number of samples is not enough to provide training in the small-scale face detection problem, for example, as shown in fig. 8.
(2) The method utilizes the generated countermeasure network to carry out high-resolution reconstruction, effectively solves the problem of insufficient input precision in the small-scale face detection problem, and is shown in the figure 9 as an example.
The above is only a preferred embodiment of the present invention, and it should be noted that it is obvious to those skilled in the art that several variations and modifications can be made without departing from the structure of the present invention, which will not affect the effect of the implementation of the present invention and the utility of the patent.
Claims (10)
1. A small-scale face detection method based on generation of a confrontation network is characterized by comprising the following steps:
s1, inputting a noise image and a small face image in real collected data into a first generation countermeasure network to generate enough small face context area samples;
s2, training a face detection network by using the small face context area samples and the small face images;
s3, acquiring a workplace image, and acquiring a possibly existing small-scale face candidate area by utilizing a potential face area network;
and S4, inputting the small-scale face candidate area into the face detection network for classification and regression to obtain a small face position regression result.
2. The method for detecting a small-scale face based on a generated confrontation network as claimed in claim 1, wherein if the small-scale face candidate region in step S4 can not be resolved to obtain a small-scale face position regression result, the following steps are performed:
s5, inputting the small-scale face candidate region into a second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S6, inputting the reconstructed face candidate region into the face detection network for classification and regression to obtain a small face position regression result.
3. The method for detecting a small-scale face based on a generative countermeasure network as claimed in claim 1, wherein the step S1 comprises the following steps:
s1-1, inputting the noise image into a generating network G1 network of the first generating countermeasure network to obtain a sample of the face context area;
s1-2, inputting the mixed set of the sample of the face context area and the face image into the discrimination network D1 network of the first generation antagonizing network to obtain the discrimination of the sample of the face context area and the face image, and optimizing the parameters of the generation network G1 network and the discrimination network D1 network under the antagonism loss adjustment of the generation network G1 network and the discrimination network D1 network.
4. The method for detecting the small-scale face based on the generation countermeasure network of claim 3, wherein the step S1-1 comprises the following steps:
s1-1-1, the noise image satisfies z-N (0,1) and the pixel size is 48x48, a series of different outputs are obtained through split operation;
s1-1-2, adding the output of split operation to different ResBlock;
and S1-1-3, performing convolution of 1x1 before the last ResBlock to perform feature fusion and channel number reduction, and extracting feature expressions generated by the G1 network to obtain a sample of the face context area.
5. The method for detecting the small-scale face based on the generation countermeasure network of claim 3, wherein the step S1-2 comprises the following steps:
s1-2-1, increasing the number of channels by convolution of 1x1 on the mixed set of the sample of the lower region of the small face and the image of the small face so as to increase the feature expression capacity;
s1-2-2, performing feature extraction through ResBlock;
s1-2-3, performing information fusion by using convolution of 1x1 and performing discrimination of the sample of the face context area and the face image in the last full connection layer.
6. The method for detecting a small-scale face based on a generative countermeasure network as claimed in claim 1, wherein the step S3 comprises the following steps:
s3-1, detecting face candidate areas under different scales by using Resnet-50 as a feature extraction network;
s3-2, performing multilayer information fusion, connecting the high-level feature map with the shallow feature map after upsampling by using the characteristic that high-level semantics guide shallow feature information, and then uniformly outputting the size by using a full connection layer;
and S3-3, establishing a small-scale face candidate region which is possibly existed by carrying out regression on the face frame based on an anchor point mechanism in the specific feature layer.
7. The method for detecting a small-scale face based on the generation of a confrontation network as claimed in claim 2, wherein said step S5 includes the following steps:
s5-1, inputting the small-scale face candidate region into the generation network G2 network of the second generation countermeasure network for super-resolution reconstruction to obtain a reconstructed face candidate region;
and S5-2, judging whether the reconstructed face candidate region is generated data or not and whether the reconstructed face candidate region is a face sample or not by utilizing the judging network D2 network of the second generation countermeasure network.
8. The method for detecting a small-scale face based on the generation of a confrontation network as claimed in claim 7, wherein the step S5-1 includes the following steps:
s5-1-1, inputting the small-scale face candidate region into the first G2 network of the second generation countermeasure network to reconstruct to obtain a sample with higher resolution;
s5-1-2, inputting the sample with higher resolution into a second generation network G2 network of the second generation countermeasure network to reconstruct the human face candidate area from higher resolution to higher resolution;
s5-1-3, training the generating network G2 network.
9. The method for detecting a small-scale face based on the generation of a confrontation network as claimed in claim 7, wherein the step S5-2 includes the following steps:
s5-2-1, judging whether the reconstructed face candidate region is the classification loss of the generated data or not and whether the reconstructed face candidate region contains the classification loss of the face or not by the judging network D2 network of the second generation countermeasure network;
s5-2-2, training the discrimination network D2 network;
s5-2-3, the discrimination network D2 network jointly optimizes according to the classification loss of generated data, the classification loss of the human face and the confrontation loss of the generation network G2 network, so that the aim of reconstructing the small-scale human face candidate region is fulfilled.
10. The small-scale face detection method based on generation of the countermeasure network according to claim 1, wherein the noise image is a randomly generated and normally distributed image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110806601.1A CN113705341A (en) | 2021-07-16 | 2021-07-16 | Small-scale face detection method based on generation countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110806601.1A CN113705341A (en) | 2021-07-16 | 2021-07-16 | Small-scale face detection method based on generation countermeasure network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113705341A true CN113705341A (en) | 2021-11-26 |
Family
ID=78648752
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110806601.1A Pending CN113705341A (en) | 2021-07-16 | 2021-07-16 | Small-scale face detection method based on generation countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113705341A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108334847A (en) * | 2018-02-06 | 2018-07-27 | 哈尔滨工业大学 | A kind of face identification method based on deep learning under real scene |
CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
CN110490076A (en) * | 2019-07-18 | 2019-11-22 | 平安科技(深圳)有限公司 | Biopsy method, device, computer equipment and storage medium |
CN110689482A (en) * | 2019-09-18 | 2020-01-14 | 中国科学技术大学 | Face super-resolution method based on supervised pixel-by-pixel generation countermeasure network |
CN110706157A (en) * | 2019-09-18 | 2020-01-17 | 中国科学技术大学 | Face super-resolution reconstruction method for generating confrontation network based on identity prior |
KR20200075063A (en) * | 2018-12-07 | 2020-06-26 | 주식회사 포스코아이씨티 | Apparatus for Extracting Face Image Based on Deep Learning |
CN112163520A (en) * | 2020-09-29 | 2021-01-01 | 广西科技大学 | MDSSD face detection method based on improved loss function |
CN113111784A (en) * | 2021-04-14 | 2021-07-13 | 山东英信计算机技术有限公司 | Face shape prediction method and system and electronic equipment |
-
2021
- 2021-07-16 CN CN202110806601.1A patent/CN113705341A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108334847A (en) * | 2018-02-06 | 2018-07-27 | 哈尔滨工业大学 | A kind of face identification method based on deep learning under real scene |
CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
KR20200075063A (en) * | 2018-12-07 | 2020-06-26 | 주식회사 포스코아이씨티 | Apparatus for Extracting Face Image Based on Deep Learning |
CN110490076A (en) * | 2019-07-18 | 2019-11-22 | 平安科技(深圳)有限公司 | Biopsy method, device, computer equipment and storage medium |
CN110689482A (en) * | 2019-09-18 | 2020-01-14 | 中国科学技术大学 | Face super-resolution method based on supervised pixel-by-pixel generation countermeasure network |
CN110706157A (en) * | 2019-09-18 | 2020-01-17 | 中国科学技术大学 | Face super-resolution reconstruction method for generating confrontation network based on identity prior |
CN112163520A (en) * | 2020-09-29 | 2021-01-01 | 广西科技大学 | MDSSD face detection method based on improved loss function |
CN113111784A (en) * | 2021-04-14 | 2021-07-13 | 山东英信计算机技术有限公司 | Face shape prediction method and system and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Huang et al. | Detection algorithm of safety helmet wearing based on deep learning | |
CN110232380B (en) | Fire night scene restoration method based on Mask R-CNN neural network | |
EP3445044B1 (en) | Video recording method, server, system, and storage medium | |
Han et al. | Deep learning-based workers safety helmet wearing detection on construction sites using multi-scale features | |
CN109325915B (en) | Super-resolution reconstruction method for low-resolution monitoring video | |
CN113435424B (en) | Method and system for identifying destroying granularity of confidential medium | |
CN116503318A (en) | Aerial insulator multi-defect detection method, system and equipment integrating CAT-BiFPN and attention mechanism | |
Qiao et al. | A weakly supervised semantic segmentation approach for damaged building extraction from postearthquake high-resolution remote-sensing images | |
CN115188066A (en) | Moving target detection system and method based on cooperative attention and multi-scale fusion | |
Chang et al. | Changes to captions: An attentive network for remote sensing change captioning | |
CN115147439A (en) | Concrete crack segmentation method and system based on deep learning and attention mechanism | |
CN114842524A (en) | Face false distinguishing method based on irregular significant pixel cluster | |
Yuan et al. | Real‐time recognition and warning of mask wearing based on improved YOLOv5 R6. 1 | |
CN114881665A (en) | Method and system for identifying electricity stealing suspected user based on target identification algorithm | |
CN116665101B (en) | Method for extracting key frames of monitoring video based on contourlet transformation | |
CN113705341A (en) | Small-scale face detection method based on generation countermeasure network | |
CN112435240B (en) | Deep vision mobile phone detection system for workers to illegally use mobile phones | |
CN115953832A (en) | Semantic decoupling-based combined action recognition method of self-attention model | |
CN115410035A (en) | Air traffic controller unsafe behavior classification method based on monitoring video | |
Chang et al. | FFA‐YOLOv7: Improved YOLOv7 Based on Feature Fusion and Attention Mechanism for Wearing Violation Detection in Substation Construction Safety | |
CN111402223B (en) | Transformer substation defect problem detection method using transformer substation video image | |
CN114170548A (en) | Oil field on-site micro-target detection method and system based on deep learning | |
CN112633142A (en) | Power transmission line violation building identification method and related device | |
Lin et al. | AC-YOLO: A Safety Helmet Detection based on YOLOX | |
Raja | Utilizing YOLO 8x Models, Deep Learning-Based Head protector Detection for Construction Workers |
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 |