Face recognition

Holographic face recognition system development

In face recognition in recent years the market application range widening, a huge market space, with the continuous improvement of face recognition technology is already relatively mature, its application field from the start of the entrance guard/attendance areas, to the current application and the financial, education, social entertainment, security terror, equipment, transportation, business areas such as intelligence, application range is gradually enlarged. On the application trend of face recognition is to transform from 2D to 3D, more accurate and more convenient. At first, facial recognition technology using 2D recognition, but as a result of 2D face recognition is vulnerable to factors such as attitude, illumination, expression, identification is not ideal, so the holographic 3 d face recognition arises at the historic moment. In comparison, the holographic 3D face recognition technology is not only high recognition rate, and on the use of convenience will be much higher than 2D face recognition. WiMi development holographic face recognition system, and can be like the human eye visual recognition, will be the first in the world of rapid, accurate human face three-dimensional identification equipment. Identify time only need one second, walk or run in the past, only on the device lapses, a look, whether it will verify its status is registered before, system for everyone's registered just 2 seconds. Don't need any direct physical contact, also do not need accurate positioning in front of the identification device, and no matter what their age or height. Holographic human face recognition system is similar to the human eye, equipped with a 3D visual system, 3D model of a face. Able to discern the geometric accuracy of up to one 5 of a millimeter, even identical twins are able to distinguish, holographic face recognition system is up and the market at present one of the most accurate biometric equipment system. Holographic application special holographic scanning technology, face recognition system using a structured light method, can quickly and accurately surfaces (depth) of the 3D model, is not affected by the ambient light, to identify more accurate. At the same time with the texture of camera, wide field of 3 d sensor can synchronization capture shape and material, the color on the surface of the object. With this system, the object of the 3D model can be obtained in the snapshot model or video mode. WiMi will provide complete solution for all areas and SDK, use the product need to pay a certain amount of technical services and after-sales service. Charge mode is divided into according to the project, according to the number of charges, according to the yearly fee, equipment related technology by shipments each charged fees, according to the video monitoring aspects' charge and charge way to profit. Can be widely applied in the entrance guard system, office furniture, business, prison, airports, customs terminal, railway station, bus station, the examination room entrance, notarization, key management, all kinds of key laboratory, office area, such as Banks and securities need accurate secure identification system management.

industry Face recognition application scenario Face recognition application scenario
The banking sector Face recognition, text recognition, live detection technology used in bank identification authentication, bill According to the charges
The financial industry, the Internet Certification than remote authentication service According to the certification service call charge
3C industry Combining with the computer, communication and consumer electronics products of science and technology integrated application of information home appliances industry According to the yearly fee
The robot industry Provide a wide range of visual recognition for robot company Most according to the technology licensing fees charged by the shipments for each
The mobile Internet industry Provide all kinds of live, picture processing, photo album APP recognition, image recognition technology Most according to the technology licensing fees charged by the shipments for each
Security industry Provide video security vendors have been supervised structured, face, face search, vehicle identification, such as population analysis software and hardware into form to provide product technical upgrade services for urban wisdom and peace city, large government projects Most according to the process monitoring video way or in accordance with the project cost
Communications industry Provide operators with recognition, text recognition, living authentication, entrance guard system, technology products such as VIP system, mobile phone CARDS real-name certification program, etc According to the project charge or operator agent take into

Product introduction

According to the user's single self-portraits, can create high-quality 3D facial model
The head model supports run-time blendshape facial animation (45 different expression)
Can change hair style, custom with the function of recolor hairstyle more (optional)
Can be attached to any body save right head. The embodiment of skeletal animation cartoon style and different rendering mode
Support user generated model integrated into Unity based, IOS and android applications

WiMi will be faster, more accurate, more reliable and more convenient way, continuing to create maximum value for customers and society

According to the forward-looking industry current situation of the development of face recognition in the industry. By 2016, the global biometric market scale at around $12.713 billion, including facial recognition roughly $2.653 billion, accounted for about 20%. Predicts 2021, the global face recognition market is expected to reach $6.37 billion, according to the expected during the compound growth rate of 17.83%. China's market size and face recognition according to the current situation of development of face recognition industry, estimate the face recognition about the size of the market accounts for about 10% of the global market. 2010-2016, our country face recognition market scale increased year by year, the average annual compound growth rate of 27%. In 2016, our country face recognition industry market size is about 1.725 billion yuan, up 27.97% year on year, growth is up 4.64% from a year earlier. As China's online identity just need and user habits, face recognition application scenario will increase further, considering the security upgrade, online identity authentication as well as the application of Internet innovation factors, the size of the market is expected to more than forecast growth in the future. Holographic capture and in addition, the development of measurement technology, face recognition algorithm based on the holographic could make up for a 2 d projection effective identification information loss problem, to face rotation, shade, and extremely similar to the traditional difficulty has the very good solution, also gradually become another important development of face recognition technology route. In face recognition industry, enterprise's target customers include the government, enterprises and individuals three categories. Among them, the main products include in the field of government oriented security products, electronic government affairs, anti-terrorism products etc. Enterprise products mainly in the field of technology solutions, terminal products, etc.; Geared to the needs of individual customers to provide products mainly include social and mobile related products. With the improvement of the holographic face library and equipment cost is reduced, the holographic face technology will be more and more applied in various industries, has a huge market space. At present in the field of face recognition is still have a few core problem in the industry, in the use of holographic facial recognition system would have to greatly improve and solve.
  • Lighting problem: in the practical application of the traditional face recognition system, will be caused by the change of environmental light detected face image in different changes in dark. Error rate is bigger, recognition success rate is low.

    Using holographic recognition methods: light face recognition system based on 3 d structure WiMi since the depth of the research learning algorithm and the structure of 3 d optical depth structure light cameras. Fault tolerance rate can reach one over one million.
  • Gestures: attitude problem is also currently a technical difficulties of facial recognition technology. At present most face recognition algorithm mainly needle columns, is positive and face image, when a pitch or left and right side under the condition of big deviation, the recognition rate of face recognition algorithm will also fell sharply.

    Using holographic recognition methods: light face recognition system based on 3 d structure WiMi since the depth of the research learning algorithm and the structure of 3 d optical depth structure light cameras. Fault tolerance rate can reach one over one million.
  • Shade problem: cooperate for the case of human face image acquisition, cover problem is a very serious problem. Glasses, hats and other accessories, the acquisition of face image is probably not complete, which affect the behind of feature extraction and recognition, even lead to the failure of face detection algorithm.

    Using holographic recognition methods: structured light projection after the light of the specific information to the surface, the camera collection. According to the object causing the change of the optical signal to calculate the information such as location and depth of the object, and then recover the whole 3 d space. Even wearing glasses, sunglasses, can quickly recognize faces.
  • Age change: as the change of the age, facial appearance is changing, especially for teenagers, this change is more obvious. For different age groups, the recognition rate of face recognition algorithm is also different.

    Ways: using holographic recognition by machine learning algorithms, each brush face computer system will be according to the features of the first photo you learn details, after every use more data and information collected.
  • A variety of expression: different caused changes in facial expression is different, in addition, different people of the same expression effect is also different, so it is difficult to use unified standard to accurately divide the various expressions of the influence of different people.

    Using holographic recognition methods: through the deep learning algorithms, each brush the face pose is impossible to exactly the same, such as deep learning system will be according to the features of each of your picture learn details, the expression of gathering more data information.
  • Background complex: when monitoring the background of the scene is more complicated, face detection rate will be reduced, thus can adapt to the complex background environment face detection algorithm is face recognition technology is very difficult to solve the problem.

    Using holographic recognition methods: structured light projected onto the face, after the rest of the background will not capture, capture the 3 d model of human face information only.

The advantage of holographic face recognition

  • High precision of recognition etc

    Registration time: 2 seconds; Recognition time: less than 1 seconds; Enrollment can reach more than thousands of database; Capacity: as many as 60 per minute

  • Faster recognition

    Registration time: 2 seconds; Recognition time: less than 1 seconds; Enrollment can reach more than thousands of database; Capacity: as many as 60 per minute

  • Reliability/security

    Registration time: 2 seconds; Recognition time: less than 1 seconds; Enrollment can reach more than thousands of database; Capacity: as many as 60 per minute

  • practical

    Registration time: 2 seconds; Recognition time: less than 1 seconds; Enrollment can reach more than thousands of database; Capacity: as many as 60 per minute

3 d face recognition compared with traditional recognition analysis
Project contrast3 d face recognition2 d face recognition
The principle of research and development technology The near infrared, structure lightVisible light
Acquisition characteristic points30000+Less than 1000
Under the weak light of recognition rate100%0%
Whether to support the nightyesno
The posture, facial expression effectHat, sunglasses, normal phone state recognitionBelt hat and sunglasses. Unable to identify call condition
Offline 1: N support identification of libraries 100,0001000-5000
Whether the identification of identical twins yesno
Requires the user to cooperation degree In any Angle of about 30 mobile identification Positive standing stationary state identification

WiMi is holographic development of face recognition

High-performance heterogeneous distributed cloud platforms, distributed deep learning platform support deep learning model and the algorithm of custom extensions, support a large number of general CPU, GPU, or CPU, GPU, mixed distributed computing.

Deep learning large-scale training system, deep learning independently by WiMi large-scale training system development, support multiple machines GPU distributed deep learning model training, support billions of parameters of the model, mass classification of hundreds of millions of categories. Industry leading memory optimization and communication optimization technology, hundreds of GPU joint training, greatly improve the speed of company training and iterative model.

Heterogeneous high-performance super calculate platform, computing power are an important force in driving the current round of artificial intelligence boom, high-performance heterogeneous computing platform, with more than 6000 pieces of high performance GPU, multiple computing cluster, the central unified storage, lightweight virtualization, researchers for the company to provide the support of a steady stream of computing power

High performance based algorithms library, heterogeneous high-performance algorithms library contains the depth all the machine learning algorithm of neural network, and mathematics and image processing algorithms. Relative to the industry open source platform libraries, bring performance improvement of 2-5 times. Support the mainstream of the cloud, personal computers, mobile client and embedded hardware platform. Support for multiple system platform, such as Linux, Android, iOS and Windows, etc.