CN108805077A - A kind of face identification system of the deep learning network based on triple loss function - Google Patents
A kind of face identification system of the deep learning network based on triple loss function Download PDFInfo
- Publication number
- CN108805077A CN108805077A CN201810592758.7A CN201810592758A CN108805077A CN 108805077 A CN108805077 A CN 108805077A CN 201810592758 A CN201810592758 A CN 201810592758A CN 108805077 A CN108805077 A CN 108805077A
- Authority
- CN
- China
- Prior art keywords
- roi
- face
- triple
- network
- convolutional neural
- 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.)
- Withdrawn
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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
A kind of face identification system of the deep learning network based on triple loss function proposed in the present invention, main contents include：Mutual correlation matches convolutional neural networks, the Heavenly Stems and Earthly Branches integrate convolutional neural networks, depth convolutional neural networks integrate and Performance Evaluation, its process is, first three samples of selection are concentrated from training data, including a static ROI (area-of-interest), a positive sample similar with static ROI and one and the negative sample of static ROI dissmilarities, these three samples is allowed to collectively constitute a triple；Then this triple is input in deep learning network and is trained, triple function (distance for the similar ROI that can further) is used in training process；Finally, similar ROI can be formed and be gathered, to achieve the purpose that recognition of face.Present invention employs triple loss functions, have higher accuracy of identification compared to traditional face identification system is played, and operation complexity is relatively low, operation efficiency is higher.
Description
Technical field
The present invention relates to field of face identification, more particularly, to a kind of deep learning net based on triple loss function
The face identification system of network.
Background technology
Recognition of face is a kind of biological identification technology that the facial feature information based on people carries out identification.Use video camera
Or camera acquires image or video flowing containing face, and automatic detect and track face in the picture, and then to detecting
Face carry out a series of relevant treatments of face, usually also referred to as Identification of Images, face recognition.Enterprise, house safety and
Management aspect, face recognition technology can be used for access control and attendance system, recognition of face antitheft door etc.；In public security, the administration of justice and criminal investigation side
Face carries out criminal in combination with face recognition technology face database and chases in the world；In E-Government and e-commerce
Use aspect can be more accurate since face identification system is using biological characteristic (rather than traditional character password)
Accomplish that party is unified in online digital identity and true identity, to greatly increase e-commerce and e-government system
Reliability；In addition, face recognition technology is also in the fields extensive use such as space flight, electric power, frontier inspection, education.However, current face
Identifying system operation complexity is higher, and operation efficiency is low, and the accuracy identified is also not high enough.
A kind of face identification system of the deep learning network based on triple loss function proposed in the present invention, first from
Training data concentrates three samples of selection, including a static ROI (area-of-interest), a positive sample similar with static ROI
This and one and the negative sample of static ROI dissmilarities, allow these three samples to collectively constitute a triple；Then by this three
Tuple is input in deep learning network and is trained, used in training process triple function (the similar ROI that can further away from
From)；Finally, similar ROI can be formed and be gathered, to achieve the purpose that recognition of face.Present invention employs triple loss function,
Have higher accuracy of identification compared to a traditional face identification system is played, and operation complexity is relatively low, operation efficiency compared with
It is high.
Invention content
Higher for current face identification system operation complexity, operation efficiency is low, and the accuracy identified
The problems such as not high enough, a kind of recognition of face system of the deep learning network based on triple loss function proposed in of the invention
System first concentrates three samples of selection, including a static ROI, a positive sample similar with static ROI and one from training data
A negative sample with static ROI dissmilarities allows these three samples to collectively constitute a triple；Then this triple is defeated
Enter into deep learning network and be trained, triple function (distance for the similar ROI that can further) is used in training process；Most
Afterwards, similar ROI can be formed and be gathered, to achieve the purpose that recognition of face.
To solve the above problems, the present invention provides a kind of face knowledge of the deep learning network based on triple loss function
Other system, main contents include：
(1) mutual correlation matching convolutional neural networks (CCM-CNN)；
(2) Heavenly Stems and Earthly Branches integrate convolutional neural networks (TBE-CNN)；
(3) depth convolutional neural networks integrate (HaarNet)；
(4) Performance Evaluation.
Wherein, the mutual correlation matches convolutional neural networks, uses matrix Hadamard product, is followed by one and connects entirely
Layer is connect, for simulating adaptive weighted cross-correlation technique；Face characterization is learnt using a kind of method optimized based on triple
Discriminate, these face characterizations be based on triple, including positive sample and negative sample video interested region (ROI) with
And corresponding static state ROI；It is non-targeted a based on static and video by generating in order to further increase the robustness of mask
The synthesis face of body ROI, it includes multinomial information to make the trim process of CCM-CNN；It is main that mutual correlation matches convolutional neural networks
Including three parts：Feature extraction, mutual correlation matching and triple loss optimization.
Further, the feature extraction is realized by feature extraction pipeline, for not sympathizing with the same object
The ROI obtained under condition carries out the extraction of distinctive feature；Feature extraction pipeline includes three sub-networks, is corresponded to respectively static, just
The face of sample and negative sample；Each sub-network includes 9 convolutional layers, is a space batch standard after each convolutional layer
Layer loses layer and line rectification function layer.
Further, mutual correlation matching, can be efficiently to spy mainly using a kind of pixel matching method
Sign mapping is compared, and is estimated it and matched similitude；Comparison process includes mainly three parts：Matrix product, full connection
Layer and Softmax layers；The method uses Feature Mapping to indicate ROI, these Feature Mappings are multiplied to triple
ROI carries out own coding, greatly reduces the complexity of comparison.
Further, triple loss optimization, is efficiently trained using a two-way triple majorized function
Network；In order to make triple loss optimization and Web-compatible, need to add additional feature extraction branch in a network；Triple
Loss can be indicated with following formula：
Wherein, Stp、StnAnd SnpSimilarity score in being matched for mutual correlation (is respectively static ROI and positive sample ROI's
Compare score, the comparison score of the comparison score and positive sample ROI and negative sample ROI of static ROI and negative sample ROI).
Wherein, the Heavenly Stems and Earthly Branches integrate convolutional neural networks, can be used for the face from overall face image and Heavenly Stems and Earthly Branches network
Complementary feature is extracted in tag block；It is artificial synthesized from static image (mainly to adopt in order to be emulated to real video data
Manually out of focus and dynamic fuzzy learns to fuzzy insensitive face characterization) fuzzy training data；TBE-CNN includes one
Core network and multiple branching networks, some utility layer of Heavenly Stems and Earthly Branches network, for being implanted into global and local information, this method
It reduces and calculates cost and effectively merged information；The output feature schematic diagram of Heavenly Stems and Earthly Branches network is together in series to connect entirely
It connects and generates final face characterization in layer.
Wherein, the depth convolutional neural networks are integrated, can efficiently learn the strong face characterization of distinctiveness to meet
Video face identifies；HaarNet contains a core network and three branching networks, and the design of these networks is for being implanted into
Facial characteristics, posture feature and other distinctive features；In addition, in order to promote discrimination, it is more that HaarNet uses one kind
The training method in stage, and additionally use a second-order statistics standard triple loss equation and obtained from changing between class in class
It wins the confidence breath；Finally, the correlation information of face ROI is implanted into a fine tuning stage, these information, which are stored in, to be logined and promoted
Identify the stage of accuracy.
Further, the multistage training method includes mainly that triple is inputted HaarNet, HaarNet
It exports result and carries out L2Standardization indicates and carries out triple loss processing, and this training method can efficiently optimize
The inner parameter of HaarNet.
Wherein, the Performance Evaluation comments this system using Cox face database (Cox Face DB)
Estimate, the facial information in Cox Face DB includes the high quality mug shot shot under controllable environment from stillcamera
With video equipment in the non-controllable low quality face-image shot；Mainly there are two aspects for Performance Evaluation：By static figure
Piece and video image compared, the assessment of computational complexity.
Further, the assessment of the computational complexity, computational complexity depends mainly on operational process, and (matching is static
ROI and video ROI) quantity, the quantity of network parameter and the number of plies；Computational complexity determines the efficiency of recognition of face.
Description of the drawings
Fig. 1 is a kind of system frame of the face identification system of the deep learning network based on triple loss function of the present invention
Frame figure.
Fig. 2 is a kind of network rack of the face identification system of the deep learning network based on triple loss function of the present invention
Composition.
Fig. 3 is a kind of training stream of face identification system of the deep learning network based on triple loss function of the present invention
Cheng Tu.
Specific implementation mode
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
It mutually combines, invention is further described in detail in the following with reference to the drawings and specific embodiments.
Fig. 1 is a kind of system frame of the face identification system of the deep learning network based on triple loss function of the present invention
Frame figure.Include mainly mutual correlation matching convolutional neural networks, the Heavenly Stems and Earthly Branches integrate convolutional neural networks, depth convolutional neural networks integrate
And Performance Evaluation.
Mutual correlation matches convolutional neural networks, uses matrix Hadamard product, is followed by a full articulamentum, is used for mould
Intend adaptive weighted cross-correlation technique；Face characterization discriminate is learnt using a kind of method optimized based on triple, these
Face characterization be based on triple, including positive sample and the video interested region (ROI) of negative sample and corresponding quiet
State ROI；In order to further increase the robustness of mask, by generating the conjunction based on static and the non-targeted individual ROI of video
At face, it includes multinomial information to make the trim process of CCM-CNN；It includes three portions that mutual correlation, which matches convolutional neural networks mainly,
Point：Feature extraction, mutual correlation matching and triple loss optimization.
The Heavenly Stems and Earthly Branches integrate convolutional neural networks, can be used for from the facial markers block of overall face image and Heavenly Stems and Earthly Branches network extracting
Complementary feature；In order to be emulated to real video data, it is artificial synthesized from static image (it is main using artificial out of focus and
Dynamic fuzzy learns to fuzzy insensitive face characterization) fuzzy training data；TBE-CNN includes core network and more
A branching networks, some utility layer of Heavenly Stems and Earthly Branches network, for being implanted into global and local information, this method, which reduces, to be calculated as
Originally and information has effectively been merged；The output feature schematic diagram of Heavenly Stems and Earthly Branches network is together in series to be generated most in full articulamentum
Whole face characterization.
Depth convolutional neural networks are integrated, can efficiently learn the strong face characterization of distinctiveness to meet video face knowledge
Not；HaarNet contains a core network and three branching networks, the designs of these networks be for be implanted into facial characteristics,
Posture feature and other distinctive features；In addition, in order to promote discrimination, HaarNet uses a kind of multistage instruction
Practice method, and additionally uses a second-order statistics standard triple loss equation and obtain information from changing between class in class；
Finally, the correlation information of face ROI is implanted into a fine tuning stage, these information, which are stored in, logins and promoted identification accurately
The stage of property.
Performance Evaluation is assessed this system, Cox Face using Cox face database (Cox Face DB)
Facial information in DB includes that the high quality mug shot shot under controllable environment from stillcamera and video equipment exist
The low quality face-image shot in the case of non-controllable；Mainly there are two aspects for Performance Evaluation：By static picture and video image
It is compared, the assessment of computational complexity.
Wherein, the assessment of computational complexity, computational complexity depend mainly on operational process and (match static ROI and video
ROI the quantity of quantity, network parameter and the number of plies)；Computational complexity determines the efficiency of recognition of face.
Fig. 2 is a kind of network rack of the face identification system of the deep learning network based on triple loss function of the present invention
Composition.Include mainly feature extraction, mutual correlation matching and triple loss optimization so that mutual correlation matches convolutional neural networks as an example
Three parts.
Feature extraction realized by feature extraction pipeline, the ROI for being obtained in varied situations to the same object into
The extraction of row distinctive feature；Feature extraction pipeline includes three sub-networks, corresponds to the face of static, positive sample and negative sample respectively
Portion；Each sub-network includes 9 convolutional layers, is a space batch index bed after each convolutional layer, loses layer and linear
Rectification function layer.
Mutual correlation matches, and mainly using a kind of pixel matching method, can efficiently be compared to Feature Mapping,
And estimates it and match similitude；Comparison process includes mainly three parts：Matrix product, full articulamentum and Softmax layers；This
Method uses Feature Mapping to indicate ROI, these Feature Mappings are multiplied to carry out own coding to the ROI of triple, greatly
The big complexity for reducing comparison.
Triple loss optimization, network is efficiently trained using a two-way triple majorized function；In order to make ternary
Group loss optimization and Web-compatible, need to add additional feature extraction branch in a network；Triple loss can be used to lower public affairs
Formula indicates：
Wherein, Stp、StnAnd SnpSimilarity score in being matched for mutual correlation (is respectively static ROI and positive sample ROI's
Compare score, the comparison score of the comparison score and positive sample ROI and negative sample ROI of static ROI and negative sample ROI).
Fig. 3 is a kind of training stream of face identification system of the deep learning network based on triple loss function of the present invention
Cheng Tu.By taking the training flow that depth convolutional neural networks integrate as an example, this training method is a kind of multistage training method, main
Include the output result progress L that triple is inputted to HaarNet, HaarNet2Standardization indicates and carries out triple loss
Processes, this training methods such as processing can efficiently optimize the inner parameter of HaarNet.
For those skilled in the art, the present invention is not limited to the details of above-described embodiment, in the essence without departing substantially from the present invention
In the case of refreshing and range, the present invention can be realized in other specific forms.In addition, those skilled in the art can be to this hair
Bright to carry out various modification and variations without departing from the spirit and scope of the present invention, these improvements and modifications also should be regarded as the present invention's
Protection domain.Therefore, the following claims are intended to be interpreted as including preferred embodiment and falls into all changes of the scope of the invention
More and change.
Claims (10)
1. a kind of face identification system of the deep learning network based on triple loss function, which is characterized in that include mainly
Mutual correlation matches convolutional neural networks (one)；The Heavenly Stems and Earthly Branches integrate convolutional neural networks (two)；Depth convolutional neural networks integrate (three)；
Performance Evaluation (four).
2. based on the mutual correlation matching convolutional neural networks (one) described in claims 1, which is characterized in that mutual correlation matching volume
Product neural network (CCM-CNN) uses matrix Hadamard product, is followed by a full articulamentum, adaptive weighted for simulating
Cross-correlation technique；Face characterization discriminate is learnt using a kind of method optimized based on triple, these face characterizations are bases
In triple, including positive sample and negative sample video interested region (ROI) and corresponding static state ROI；In order into
One step improves the robustness of mask, by generating based on static and the non-targeted individual ROI of video synthesis face, makes CCM-
The trim process of CNN includes multinomial information；It includes three parts that mutual correlation, which matches convolutional neural networks mainly,：Feature extraction,
Mutual correlation matches and triple loss optimization.
3. based on the feature extraction described in claims 2, which is characterized in that realized by feature extraction pipeline, for same
The ROI that one object obtains in varied situations carries out the extraction of distinctive feature；Feature extraction pipeline includes three sub-networks,
The face of static, positive sample and negative sample is corresponded to respectively；Each sub-network includes 9 convolutional layers, after each convolutional layer
It is a space batch index bed, loses layer and line rectification function layer.
4. based on the mutual correlation matching described in claims 2, which is characterized in that mutual correlation matches mainly using a kind of picture
Plain matching process can efficiently compare Feature Mapping, and estimate it and match similitude；Comparison process includes mainly
Three parts：Matrix product, full articulamentum and Softmax layers；The method uses Feature Mapping to indicate ROI, by these spies
Sign mapping is multiplied to carry out own coding to the ROI of triple, greatly reduces the complexity of comparison.
5. losing optimization based on the triple described in claims 2, which is characterized in that optimize letter using a two-way triple
It counts efficiently to train network；In order to make triple loss optimization and Web-compatible, need to add additional feature in a network
Extracting branch；Triple loss can be indicated with following formula：
Wherein, Stp、StnAnd SnpSimilarity score in being matched for mutual correlation (is respectively the comparison of static ROI and positive sample ROI
The comparison score of score, the comparison score and positive sample ROI and negative sample ROI of static ROI and negative sample ROI).
6. integrating convolutional neural networks (two) based on the Heavenly Stems and Earthly Branches described in claims 1, which is characterized in that the Heavenly Stems and Earthly Branches integrate convolution god
It can be used for extracting complementary feature from the facial markers block of overall face image and Heavenly Stems and Earthly Branches network through network (TBE-CNN)；For
Real video data emulated, it is artificial synthesized from static image (main out of focus to be learned with dynamic fuzzy using artificial
Practise to fuzzy insensitive face characterization) obscure training data；TBE-CNN includes a core network and multiple branching networks,
Some utility layer of Heavenly Stems and Earthly Branches network, for being implanted into global and local information, this method reduce calculate cost and effectively
Information is merged in ground；The output feature schematic diagram of Heavenly Stems and Earthly Branches network is together in series to generate final facial table in full articulamentum
Sign.
7. integrating (three) based on the depth convolutional neural networks described in claims 1, which is characterized in that depth convolutional Neural net
Network, which integrates (HaarNet), can efficiently learn the strong face characterization of distinctiveness to meet video face identification；HaarNet is contained
One core network and three branching networks, the designs of these networks be for be implanted into facial characteristics, posture feature and other
Distinctive feature；In addition, in order to promote discrimination, HaarNet uses a kind of multistage training method, and also uses
One second-order statistics standard triple loss equation obtains information from changing between class in class；Finally, rank is finely tuned at one
The correlation information of implantation face ROI, these information are stored in the stage for logining and being promoted identification accuracy in section.
8. based on the multistage training method described in claims 7, which is characterized in that main includes inputting triple
The output result of HaarNet, HaarNet carry out L2Standardization indicates and carries out triple loss processing, this training method energy
Enough inner parameters for efficiently optimizing HaarNet.
9. based on the Performance Evaluation (four) described in claims 1, which is characterized in that use Cox face database (Cox
Face DB) this system to be assessed, the facial information in Cox Face DB includes from stillcamera in controllable environment
The high quality mug shot and video equipment of lower shooting are in the non-controllable low quality face-image shot；Performance Evaluation master
It will be there are two aspect：Static picture and video image are compared, the assessment of computational complexity.
10. the assessment based on the computational complexity described in claims 9, which is characterized in that computational complexity depends mainly on
Quantity, the quantity of network parameter and the number of plies of operational process (matching static ROI and video ROI)；Computational complexity determines people
The efficiency of face identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810592758.7A CN108805077A (en) | 2018-06-11 | 2018-06-11 | A kind of face identification system of the deep learning network based on triple loss function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810592758.7A CN108805077A (en) | 2018-06-11 | 2018-06-11 | A kind of face identification system of the deep learning network based on triple loss function |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108805077A true CN108805077A (en) | 2018-11-13 |
Family
ID=64088221
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810592758.7A Withdrawn CN108805077A (en) | 2018-06-11 | 2018-06-11 | A kind of face identification system of the deep learning network based on triple loss function |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108805077A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110096580A (en) * | 2019-04-24 | 2019-08-06 | 北京百度网讯科技有限公司 | A kind of FAQ dialogue method, device and electronic equipment |
| CN110472570A (en) * | 2019-08-14 | 2019-11-19 | 旭辉卓越健康信息科技有限公司 | A kind of recognition of face multipath deep neural network method based on adaptive weighting |
| CN110866140A (en) * | 2019-11-26 | 2020-03-06 | 腾讯科技（深圳）有限公司 | Image feature extraction model training method, image searching method and computer equipment |
| CN111274445A (en) * | 2020-01-20 | 2020-06-12 | 山东建筑大学 | Similar video content retrieval method and system based on triple deep learning |
| CN111325051A (en) * | 2018-12-13 | 2020-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
| CN111860696A (en) * | 2020-08-03 | 2020-10-30 | 上海高德威智能交通***有限公司 | Character registration method and device, electronic equipment and storage medium |
| CN112949384A (en) * | 2021-01-23 | 2021-06-11 | 西北工业大学 | Remote sensing image scene classification method based on antagonistic feature extraction |
| CN113435383A (en) * | 2021-07-07 | 2021-09-24 | 中国人民解放军国防科技大学 | Remote sensing airplane target classification method and device based on double triple pseudo-twin framework |
| CN113538075A (en) * | 2020-04-14 | 2021-10-22 | 阿里巴巴集团控股有限公司 | Data processing method, model training method, device and equipment |
| CN113850243A (en) * | 2021-11-29 | 2021-12-28 | 北京的卢深视科技有限公司 | Model training method, face recognition method, electronic device and storage medium |
| CN114972740A (en) * | 2022-07-29 | 2022-08-30 | 上海鹰觉科技有限公司 | Automatic ship sample collection method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080285968A1 (en) * | 2007-05-14 | 2008-11-20 | Powergate Optical Inc. | Compact camera module package structure |
| CN102754520A (en) * | 2010-02-11 | 2012-10-24 | 三星电子株式会社 | Method and apparatus for providing information in mobile terminal |
-
2018
- 2018-06-11 CN CN201810592758.7A patent/CN108805077A/en not_active Withdrawn
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080285968A1 (en) * | 2007-05-14 | 2008-11-20 | Powergate Optical Inc. | Compact camera module package structure |
| CN102754520A (en) * | 2010-02-11 | 2012-10-24 | 三星电子株式会社 | Method and apparatus for providing information in mobile terminal |
Non-Patent Citations (1)
| Title |
|---|
| MOSTAFA PARCHAMI ET AL: ""CNNs with cross-correlation matching for face recognition in video surveillance using a single training sample per person"", 《IEEE:HTTPS://IEEEXPLORE.IEEE.ORG/STAMP/STAMP.JSP?TP=&ARNUMBER=8078554》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111325051B (en) * | 2018-12-13 | 2023-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
| CN111325051A (en) * | 2018-12-13 | 2020-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
| CN110096580A (en) * | 2019-04-24 | 2019-08-06 | 北京百度网讯科技有限公司 | A kind of FAQ dialogue method, device and electronic equipment |
| CN110472570A (en) * | 2019-08-14 | 2019-11-19 | 旭辉卓越健康信息科技有限公司 | A kind of recognition of face multipath deep neural network method based on adaptive weighting |
| CN110866140A (en) * | 2019-11-26 | 2020-03-06 | 腾讯科技（深圳）有限公司 | Image feature extraction model training method, image searching method and computer equipment |
| CN110866140B (en) * | 2019-11-26 | 2024-02-02 | 腾讯科技（深圳）有限公司 | Image feature extraction model training method, image searching method and computer equipment |
| CN111274445A (en) * | 2020-01-20 | 2020-06-12 | 山东建筑大学 | Similar video content retrieval method and system based on triple deep learning |
| CN113538075A (en) * | 2020-04-14 | 2021-10-22 | 阿里巴巴集团控股有限公司 | Data processing method, model training method, device and equipment |
| CN111860696A (en) * | 2020-08-03 | 2020-10-30 | 上海高德威智能交通***有限公司 | Character registration method and device, electronic equipment and storage medium |
| CN112949384A (en) * | 2021-01-23 | 2021-06-11 | 西北工业大学 | Remote sensing image scene classification method based on antagonistic feature extraction |
| CN112949384B (en) * | 2021-01-23 | 2024-03-08 | 西北工业大学 | Remote sensing image scene classification method based on antagonistic feature extraction |
| CN113435383A (en) * | 2021-07-07 | 2021-09-24 | 中国人民解放军国防科技大学 | Remote sensing airplane target classification method and device based on double triple pseudo-twin framework |
| CN113850243A (en) * | 2021-11-29 | 2021-12-28 | 北京的卢深视科技有限公司 | Model training method, face recognition method, electronic device and storage medium |
| CN114972740A (en) * | 2022-07-29 | 2022-08-30 | 上海鹰觉科技有限公司 | Automatic ship sample collection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108805077A (en) | A kind of face identification system of the deep learning network based on triple loss function | |
| CN113378632B (en) | Pseudo-label optimization-based unsupervised domain adaptive pedestrian re-identification method | |
| CN110163236B (en) | Model training method and device, storage medium and electronic device | |
| Sultana et al. | Social behavioral information fusion in multimodal biometrics | |
| CN109447115A (en) | Zero sample classification method of fine granularity based on multilayer semanteme supervised attention model | |
| CN109359541A (en) | A kind of sketch face identification method based on depth migration study | |
| Wan et al. | Auxiliary demographic information assisted age estimation with cascaded structure | |
| CN106815566A (en) | A kind of face retrieval method based on multitask convolutional neural networks | |
| CN108345837A (en) | A kind of pedestrian's recognition methods again based on the study of human region alignmentization feature representation | |
| CN110598543B (en) | Model training method based on attribute mining and reasoning and pedestrian re-identification method | |
| CN106022317A (en) | Face identification method and apparatus | |
| CN109815826A (en) | The generation method and device of face character model | |
| CN108960080A (en) | Based on Initiative Defense image to the face identification method of attack resistance | |
| CN105095870A (en) | Pedestrian re-recognition method based on transfer learning | |
| CN112801054B (en) | Face recognition model processing method, face recognition method and device | |
| Nguyen et al. | Generating master faces for use in performing wolf attacks on face recognition systems | |
| Hu et al. | Toward driver face recognition in the intelligent traffic monitoring systems | |
| Gao et al. | Automatic facial attractiveness prediction by deep multi-task learning | |
| CN114662497A (en) | False news detection method based on cooperative neural network | |
| Kumar Verma et al. | Generative model for zero-shot sketch-based image retrieval | |
| CN111126307A (en) | Small sample face recognition method of joint sparse representation neural network | |
| CN110443232A (en) | Method for processing video frequency and relevant apparatus, image processing method and relevant apparatus | |
| CN107220598A (en) | Iris Texture Classification based on deep learning feature and Fisher Vector encoding models | |
| CN109271546A (en) | The foundation of image retrieval Feature Selection Model, Database and search method | |
| Zhong et al. | Palmprint and dorsal hand vein dualmodal biometrics |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181113 |