CN113963420B - A 3D face recognition method based on deep learning - Google Patents
A 3D face recognition method based on deep learning Download PDFInfo
- Publication number
- CN113963420B CN113963420B CN202111326034.6A CN202111326034A CN113963420B CN 113963420 B CN113963420 B CN 113963420B CN 202111326034 A CN202111326034 A CN 202111326034A CN 113963420 B CN113963420 B CN 113963420B
- Authority
- CN
- China
- Prior art keywords
- face
- dimensional
- color
- training sample
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
一种基于深度学习的三维人脸识别方法，包括：构建人脸三维深度学习网络，计算每个训练样本的高频‑深度映射图，然后将每个训练样本的彩色人脸二维图像和高频‑深度映射图同时输入人脸三维深度学习网络中训练；使用彩色摄像头拍摄获得待测人脸的彩色人脸二维图像，然后将待测人脸的彩色人脸二维图像输入训练后的人脸三维深度学习网络，输出获得待测人脸的三维点云坐标；根据待测人脸的彩色人脸二维图像和三维点云坐标，计算三维人脸特征向量，然后将待测人脸的三维人脸特征向量和注册库中已注册人脸的三维人脸特征向量进行对比，从而识别待测人脸的人员信息。本发明属于信息技术领域，能充分利用人脸的深度信息来有效提高人脸识别结果的可靠性。
A three-dimensional face recognition method based on deep learning includes: constructing a three-dimensional face deep learning network, calculating the high-frequency-depth map of each training sample, and then simultaneously inputting the color face two-dimensional image and the high-frequency-depth map of each training sample into the face three-dimensional deep learning network for training; using a color camera to obtain a color face two-dimensional image of a face to be tested, and then inputting the color face two-dimensional image of the face to be tested into the trained face three-dimensional deep learning network, and outputting the three-dimensional point cloud coordinates of the face to be tested; according to the color face two-dimensional image and the three-dimensional point cloud coordinates of the face to be tested, calculating a three-dimensional face feature vector, and then comparing the three-dimensional face feature vector of the face to be tested with the three-dimensional face feature vector of the registered face in the registration library, so as to identify the personnel information of the face to be tested. The present invention belongs to the field of information technology, and can make full use of the depth information of the face to effectively improve the reliability of the face recognition result.
Description
Technical Field
The invention relates to a three-dimensional face recognition method based on deep learning, and belongs to the technical field of information.
Background
The face recognition is a computer technology for carrying out identity authentication by analyzing face visual characteristic information, belongs to non-contact recognition, has high concealment, and is widely applied to the fields with high safety factor requirements such as office sites, construction sites and the like.
In recent years, with the development of deep learning technology, face recognition technology using neural networks is mature, and related applications are falling to the ground, but due to the loss of depth information in the face imaging process, reliability of all recognition results is still insufficient.
Therefore, how to fully utilize depth information of a face to effectively improve reliability of a face recognition result has become one of the technical problems to be solved in the prior art.
Disclosure of Invention
Accordingly, the present invention aims to provide a three-dimensional face recognition method based on deep learning, which can fully utilize depth information of a face to effectively improve reliability of a face recognition result.
In order to achieve the above purpose, the present invention provides a three-dimensional face recognition method based on deep learning, comprising:
Step one, constructing a three-dimensional deep learning network of a human face, calculating a high-frequency-depth mapping diagram of each training sample, and then inputting a color human face two-dimensional image and the high-frequency-depth mapping diagram of each training sample into the three-dimensional deep learning network of the human face at the same time for training;
Shooting by using a color camera to obtain a color face two-dimensional image of the face to be detected, inputting the color face two-dimensional image of the face to be detected into a trained face three-dimensional deep learning network, and outputting to obtain three-dimensional point cloud coordinates of the face to be detected;
Step three, calculating three-dimensional face feature vectors according to the two-dimensional images of the color faces and the three-dimensional point cloud coordinates of the faces to be detected, comparing the three-dimensional face feature vectors of the faces to be detected with the three-dimensional face feature vectors of the registered faces in the registration library, thereby identifying personnel information of the faces to be detected,
Calculating a high frequency-depth map of any training sample X, further comprising:
step 11, carrying out Fourier spectrum transformation on the color face two-dimensional image of the training sample X, thereby obtaining a spectrogram of the training sample X;
Step 12, extracting a plurality of key points from the color face two-dimensional image of the training sample X, then reading the frequency spectrum value of each key point from the spectrogram of the training sample X, and calculating the truncated value D 0,D0 of the high-pass filter function as the average value of the frequency spectrum values of all the key points according to the frequency spectrum value;
Step 13, setting a high-pass filter function, and passing the spectrogram of the training sample X through the high-pass filter function to obtain a filtered high-frequency chart, wherein the high-pass filter function is set as follows: Wherein D (u, v) is a frequency spectrum value of a coordinate (u, v) on a spectrogram of the training sample X, n is an order constant, 2 or 4 is taken, and H (u, v) is a frequency spectrum value obtained after filtering;
step 14, performing inverse Fourier transform on the high-frequency image of the training sample X, so as to obtain a high-frequency color face two-dimensional image;
and 15, comparing the brightness value of each point in the high-frequency color face two-dimensional image with a threshold value, replacing the brightness value higher than the threshold value with a depth value, and obtaining the high-frequency color face two-dimensional image which is the high-frequency-depth map of the training sample X after all the points are compared.
Compared with the prior art, the face detection method has the beneficial effects that the face depth information is added into the training process of the neural network, the defect of the depth information is made up, so that the reliability of the face recognition result is effectively improved, only one camera is adopted, only a single image of the detected face is required to be acquired, and the detection speed, the passing rate and the anti-counterfeiting rate of the face living body are greatly improved by combining the technologies of deep learning, machine learning and the like.
Drawings
Fig. 1 is a flowchart of a three-dimensional face recognition method based on deep learning.
Fig. 2 is a specific flowchart of calculating a high frequency-depth map of any training sample X.
Fig. 3 is a specific flowchart of calculating a three-dimensional face feature vector according to the two-dimensional image of the color face and the three-dimensional point cloud coordinates of the face to be measured in the step three of fig. 1.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings, in order to make the objects, technical solutions and advantages of the present invention more apparent.
As shown in fig. 1, the three-dimensional face recognition method based on deep learning of the present invention includes:
Step one, constructing a three-dimensional deep learning network of a human face, calculating a high-frequency-depth mapping diagram of each training sample, and then inputting a color human face two-dimensional image and the high-frequency-depth mapping diagram of each training sample into the three-dimensional deep learning network of the human face at the same time for training;
Shooting by using a color camera to obtain a color face two-dimensional image of the face to be detected, inputting the color face two-dimensional image of the face to be detected into a trained face three-dimensional deep learning network, and outputting to obtain three-dimensional point cloud coordinates of the face to be detected;
And thirdly, calculating a three-dimensional face feature vector according to the two-dimensional image of the color face and the three-dimensional point cloud coordinates of the face to be detected, and comparing the three-dimensional face feature vector of the face to be detected with the three-dimensional face feature vector of the registered face in the registry, so as to identify the personnel information of the face to be detected.
When registering a face, the 3d sensor can be used for accurately recording three-dimensional information of the face, generating a three-dimensional face feature vector, and finally storing the three-dimensional face feature vector of the registered face into a registration library.
The specific structure of the face three-dimensional deep learning network in the first step is constructed as follows:
the method is realized based on tensorflow frames and adopts an encoding-decoding (namely an encoder-decoder) structure, wherein the encoder part comprises 1 convolution layer and 7 residual layers, so that an input 250 x 3 face two-dimensional image can be converted into a 16 x 1024 feature image, 9 deconvolution layers are arranged in the decoder part, the feature image can be converted into a 250 x 1 three-dimensional depth image, the three-dimensional depth image is converted into a three-dimensional point cloud through a fixed conversion relation between the three-dimensional depth image and point cloud, the sizes of the convolution layers and deconvolution layer cores are 4, and an activation function is Softplus. Thus, when a 250×250×3 color face two-dimensional image is input, the output of the face three-dimensional deep learning network is 250×250 three-dimensional point cloud coordinates (62500 total).
It is worth mentioning that, in training the three-dimensional deep learning network of human face, the invention can be through calculating the high frequency-depth map of each training sample, and input the two-dimensional image of color human face and high frequency-depth map of each training sample into the three-dimensional deep learning network of human face at the same time to train, thus make the weight of the general characteristic (namely the characteristic that is relatively common in human face and difficult to be used for carrying on the recognition) in the human face image reduce in the model parameter obtained after training, the weight of the individual characteristic (namely the characteristic that is relatively individual in human face and helpful for accurate recognition) improves, improve the accuracy of human face obviously. As shown in fig. 2, the calculation of the high frequency-depth map of any training sample X may further include:
step 11, carrying out Fourier spectrum transformation on the color face two-dimensional image of the training sample X, thereby obtaining a spectrogram of the training sample X;
step 12, extracting a plurality of key points from the color face two-dimensional image of the training sample X, wherein the key points can be eyebrows, eyes, noses, mouths, facial contours and the like, then reading the frequency spectrum value of each key point from the frequency spectrum diagram of the training sample X, and calculating a cutoff value D 0,D0 of a high-pass filter function according to the frequency spectrum value of each key point;
Step 13, setting a high-pass filter function, and passing the spectrogram of the training sample X through the high-pass filter function to obtain a filtered high-frequency chart, wherein the high-pass filter function is set as follows: Wherein D (u, v) is the frequency spectrum value of the coordinate (u, v) on the spectrogram of the training sample X, n is the order constant, 2 or 4 is taken, H (u, v) is the frequency spectrum value obtained after filtering, thus the low-frequency information can be effectively removed, and the required high-frequency information can be obtained;
step 14, performing inverse Fourier transform on the high-frequency image of the training sample X, so as to obtain a high-frequency color face two-dimensional image;
And 15, comparing the brightness value of each point in the high-frequency color face two-dimensional image with a threshold value, and replacing the brightness value higher than the threshold value with a depth value, namely judging whether the brightness value of each point is higher than the threshold value, if so, reading the three-dimensional coordinates of the point from the three-dimensional point cloud of the training sample X, converting the read three-dimensional coordinates into the depth value, replacing the brightness value of the point with the depth value, and if not, continuing judging the next point, wherein the obtained high-frequency color face two-dimensional image is the high-frequency-depth map of the training sample X after all the points are compared. The three-dimensional point cloud of the training sample X may be obtained by collecting the training sample using a depth camera, and the threshold may be set according to actual service needs, for example, 20.
As shown in fig. 3, in the third step, according to the two-dimensional image of the color face and the three-dimensional point cloud coordinates of the face to be detected, a three-dimensional face feature vector is calculated, which may further include:
Step 31, performing face detection, cutting and alignment on a two-dimensional image of a face to be detected, converting the two-dimensional image into a floating point matrix, and then calculating a corresponding two-dimensional image feature vector, wherein the two-dimensional image feature vector is a one-dimensional vector with a size of 512;
Step 32, according to the three-dimensional point cloud coordinates of the face to be detected, a Cartesian xyz rectangular coordinate system is established by taking the nose tip as a coordinate origin, taking a binocular connecting line as a transverse axis direction x and taking a connecting line between the nose tip and the lips as a longitudinal axis direction y, so that a three-dimensional point cloud matrix of the face is obtained, and then a corresponding three-dimensional point cloud feature vector is calculated, wherein the three-dimensional point cloud feature vector is a three-dimensional vector with the size of 3 x 512;
And 33, setting weight factors, namely fusing the two-dimensional image feature vector and the three-dimensional point cloud feature vector, wherein the feature vector obtained after fusion is the three-dimensional face feature vector, and the weight factors of the two-dimensional face feature, the x-direction feature, the y-direction feature and the z-direction feature can be respectively set to be 0.5, 0.17 and 0.16.
In the third step, the three-dimensional face feature vector of the face to be detected is compared with the three-dimensional face feature vector of any registered face in the registration library, and the method may further include:
Step A1, calculating cosine similarity between a three-dimensional face feature vector A of a face to be detected and a three-dimensional face feature vector B of a registered face, wherein a calculation formula can be as follows: Wherein a·b represents the dot product of vectors a and B, |a| 2 represents the L2 norm of vector a, |b|| 2 represents the L2 norm of vector B;
And A2, calculating cosine distances between the face to be detected and the registered face, wherein dist (A, B) =1-cos (A, B), and if the cosine distances are larger, the similarity between the two faces is lower, and judging whether the face to be detected is the registered face according to the cosine distances.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather to enable any modification, equivalent replacement, improvement or the like to be made within the spirit and principles of the invention.
Claims (6)
1. The three-dimensional face recognition method based on deep learning is characterized by comprising the following steps of:
Step one, constructing a three-dimensional deep learning network of a human face, calculating a high-frequency-depth mapping diagram of each training sample, and then inputting a color human face two-dimensional image and the high-frequency-depth mapping diagram of each training sample into the three-dimensional deep learning network of the human face at the same time for training;
Shooting by using a color camera to obtain a color face two-dimensional image of the face to be detected, inputting the color face two-dimensional image of the face to be detected into a trained face three-dimensional deep learning network, and outputting to obtain three-dimensional point cloud coordinates of the face to be detected;
Step three, calculating three-dimensional face feature vectors according to the two-dimensional images of the color faces and the three-dimensional point cloud coordinates of the faces to be detected, comparing the three-dimensional face feature vectors of the faces to be detected with the three-dimensional face feature vectors of the registered faces in the registration library, thereby identifying personnel information of the faces to be detected,
Calculating a high frequency-depth map of any training sample X, further comprising:
step 11, carrying out Fourier spectrum transformation on the color face two-dimensional image of the training sample X, thereby obtaining a spectrogram of the training sample X;
Step 12, extracting a plurality of key points from the color face two-dimensional image of the training sample X, then reading the frequency spectrum value of each key point from the spectrogram of the training sample X, and calculating the truncated value D 0,D0 of the high-pass filter function as the average value of the frequency spectrum values of all the key points according to the frequency spectrum value;
Step 13, setting a high-pass filter function, and passing the spectrogram of the training sample X through the high-pass filter function to obtain a filtered high-frequency chart, wherein the high-pass filter function is set as follows: Wherein D (u, v) is a frequency spectrum value of a coordinate (u, v) on a spectrogram of the training sample X, n is an order constant, 2 or 4 is taken, and H (u, v) is a frequency spectrum value obtained after filtering;
step 14, performing inverse Fourier transform on the high-frequency image of the training sample X, so as to obtain a high-frequency color face two-dimensional image;
and 15, comparing the brightness value of each point in the high-frequency color face two-dimensional image with a threshold value, replacing the brightness value higher than the threshold value with a depth value, and obtaining the high-frequency color face two-dimensional image which is the high-frequency-depth map of the training sample X after all the points are compared.
2. The method of claim 1, wherein the specific structure of the three-dimensional deep learning network of the face in the step one is constructed as follows:
The method is realized based on tensorflow frames and adopts a coding-decoding encoder-decoder structure, wherein the encoder part comprises 1 convolution layer and 7 residual layers, so that an input 250 x3 face two-dimensional image is converted into a 16 x 1024 feature image, 9 deconvolution layers are arranged in the decoder part, so that the feature image is converted into a 250 x 1 three-dimensional depth image, the three-dimensional depth image is converted into a three-dimensional point cloud for output through a fixed conversion relation between the three-dimensional depth image and point cloud, the sizes of the convolution layer and the deconvolution layer core are 4, and an activation function adopts Softplus.
3. The method of claim 1, wherein step 15 further comprises:
judging whether the brightness value of each point is higher than a threshold value, if so, reading the three-dimensional coordinates of the point from the three-dimensional point cloud of the training sample X, converting the read three-dimensional coordinates into a depth value, replacing the brightness value of the point with the depth value, and if not, continuing to judge the next point.
4. The method according to claim 1, wherein in the third step, a three-dimensional face feature vector is calculated according to a two-dimensional image of a color face and three-dimensional point cloud coordinates of the face to be measured, and further comprising:
Step 31, performing face detection, cutting and alignment on a color face two-dimensional image of a face to be detected, converting the color face two-dimensional image into a floating point matrix, and then calculating a corresponding two-dimensional image feature vector, wherein the two-dimensional image feature vector is a one-dimensional vector with a size of 512;
Step 32, according to the three-dimensional point cloud coordinates of the face to be detected, a Cartesian xyz rectangular coordinate system is established by taking the nose tip as a coordinate origin, taking a binocular connecting line as a transverse axis direction x and taking a connecting line between the nose tip and the lips as a longitudinal axis direction y, so that a three-dimensional point cloud matrix of the face is obtained, and then a corresponding three-dimensional point cloud feature vector is calculated, wherein the three-dimensional point cloud feature vector is a three-dimensional vector with the size of 3 x 512;
and 33, setting a weight factor, and fusing the two-dimensional image feature vector and the three-dimensional point cloud feature vector, wherein the feature vector obtained after fusion is the three-dimensional face feature vector.
5. The method of claim 4, wherein the weighting factors of the two-dimensional face feature, the x-direction feature, the y-direction feature, and the z-direction feature are set to 0.5, 0.17, and 0.16, respectively, in step 33.
6. The method according to claim 1, wherein in the third step, the three-dimensional face feature vector of the face to be detected is compared with the three-dimensional face feature vector of any registered face in the registry, and further comprising:
step A1, calculating cosine similarity between a three-dimensional face feature vector A of a face to be detected and a three-dimensional face feature vector B of a registered face, wherein the calculation formula is as follows: Wherein A.B represents the dot product of vectors A and B, II A II 2 represents the L2 norm of vector A, and II B II 2 represents the L2 norm of vector B;
And A2, calculating cosine distances between the face to be detected and the registered face, wherein dist (A, B) =1-cos (A, B), and judging whether the face to be detected is the registered face according to the cosine distances.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111326034.6A CN113963420B (en) | 2021-11-10 | 2021-11-10 | A 3D face recognition method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111326034.6A CN113963420B (en) | 2021-11-10 | 2021-11-10 | A 3D face recognition method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113963420A CN113963420A (en) | 2022-01-21 |
| CN113963420B true CN113963420B (en) | 2025-01-07 |
Family
ID=79469931
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111326034.6A Active CN113963420B (en) | 2021-11-10 | 2021-11-10 | A 3D face recognition method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113963420B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115546835A (en) * | 2022-10-14 | 2022-12-30 | 广东华穗科技有限公司 | An AI-based mask wearing detection method and system |
| CN116682181A (en) * | 2023-05-31 | 2023-09-01 | 四川新网银行股份有限公司 | Vibration signal-based living body detection method, device and storage medium |
| GB202312687D0 (en) * | 2023-08-18 | 2023-10-04 | Rewire Holding Ltd | System and method |
| CN118172820B (en) * | 2024-04-15 | 2024-09-20 | 中国矿业大学 | Three-dimensional face recognition method applied to point cloud |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077796A (en) * | 2013-03-28 | 2014-10-01 | 北京京东方光电科技有限公司 | Method and device for converting two-dimensional image into three-dimensional image |
| CN108133459A (en) * | 2016-12-01 | 2018-06-08 | 北京大学 | Depth map Enhancement Method and depth map intensifier |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104504410A (en) * | 2015-01-07 | 2015-04-08 | 深圳市唯特视科技有限公司 | Three-dimensional face recognition device and method based on three-dimensional point cloud |
| CN105956582B (en) * | 2016-06-24 | 2019-07-30 | 深圳市唯特视科技有限公司 | A kind of face identification system based on three-dimensional data |
| CN109670487A (en) * | 2019-01-30 | 2019-04-23 | 汉王科技股份有限公司 | A kind of face identification method, device and electronic equipment |
| CN112966670A (en) * | 2021-04-08 | 2021-06-15 | 北京的卢深视科技有限公司 | Face recognition method, electronic device and storage medium |
| CN113239885A (en) * | 2021-06-04 | 2021-08-10 | 新大陆数字技术股份有限公司 | Face detection and recognition method and system |
-
2021
- 2021-11-10 CN CN202111326034.6A patent/CN113963420B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077796A (en) * | 2013-03-28 | 2014-10-01 | 北京京东方光电科技有限公司 | Method and device for converting two-dimensional image into three-dimensional image |
| CN108133459A (en) * | 2016-12-01 | 2018-06-08 | 北京大学 | Depth map Enhancement Method and depth map intensifier |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113963420A (en) | 2022-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113963420B (en) | A 3D face recognition method based on deep learning | |
| CN109978871B (en) | Fiber tract screening method combining probabilistic and deterministic fiber tract tracing | |
| CN115797970B (en) | Dense pedestrian target detection method and system based on YOLOv5 model | |
| CN116152697B (en) | Three-dimensional model measuring method and related device for concrete structure cracks | |
| CN108154066B (en) | A 3D Object Recognition Method Based on Curvature Feature Recurrent Neural Network | |
| CN110874587B (en) | A facial feature parameter extraction system | |
| CN109544523A (en) | Quality of human face image evaluation method and device based on more attribute face alignments | |
| CN111639629A (en) | Pig weight measuring method and device based on image processing and storage medium | |
| CN104834905A (en) | Facial image identification simulation system and method | |
| CN114170686A (en) | Elbow bending behavior detection method based on human body key points | |
| CN111127417A (en) | Soft package coil stock printing defect detection method based on SIFT feature matching and improved SSD algorithm | |
| CN111833307A (en) | Method and device for diagnosing cracks in wind turbine blades | |
| WO2024255161A1 (en) | Facial anti-counterfeiting method and apparatus, and storage medium | |
| CN106327534A (en) | Tire inner wall texture identification method based on locating block | |
| CN114842078A (en) | Dual-channel satellite attitude estimation network based on deep learning | |
| CN120124032A (en) | Intelligent lock unlocking method and system based on face image processing | |
| CN118570852A (en) | Face aging index evaluation system based on deep learning | |
| CN118089669A (en) | Topography mapping system and method based on aviation mapping technology | |
| CN100561500C (en) | Face Pose Recognition Method Based on Restricted Boltzmann Machine Neural Network | |
| CN117496570A (en) | Face recognition method and system based on multi-scale convolution | |
| CN112489197B (en) | A single-view 3D reconstruction system and method based on adversarial training prior learning | |
| CN113870351B (en) | Indoor large scene pedestrian fingerprint positioning method based on monocular vision | |
| Arunkumar et al. | Estimation of vehicle distance based on feature points using monocular vision | |
| CN109886091A (en) | 3D facial expression recognition method based on weighted local curl pattern | |
| CN119888410A (en) | Multi-view collaborative cross-modal 3D perception model training method based on data screening optimization |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 310013 4th floor, No.398 Wensan Road, Xihu District, Hangzhou City, Zhejiang Province Applicant after: Xinxun Digital Technology (Hangzhou) Co.,Ltd. Address before: 310013 4th floor, No.398 Wensan Road, Xihu District, Hangzhou City, Zhejiang Province Applicant before: EB Information Technology Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |