Face recognition

Development of holographic face recognition system


   In the past few years, the application scope of face recognition market has been expanded gradually, which forms a huge market space. With the continuous improvement of face recognition technology, the application area has been relatively mature. The application area has been gradually expanded from the field of access control/attendance to the current application in finance, security anti-terrorism, education, social entertainment, equipment, traffic, intelligent business and so on.


   The application trend of face recognition will also change from 2D to 3D, which is more accurate and convenient. At first, 2D recognition technology is used in face recognition, but because it is easily affected by postures, illumination, expression and other factors, the recognition rate is not good enough, which made the holographic 3D face recognition emerges as the times require. In contrast, holographic 3D face recognition technology not only has a high recognition rate, but also is much easier to use than 2D face recognition.


   The holographic face recognition system developed by WIMI HOLOGRAM CLOUD can recognize face as intuitively as human eyes, which will be the first 3D face recognition equipment with efficiency and accuracy in the world. It takes only a second to authenticate, and even to run over with just one look on the device when you walk, it will verify whether its identity has been registered or not, and the registration time for everyone by the system is only 2 seconds. There is no need for any direct physical contact, nor does it need to be accurately positioned in front of the identification device, regardless of their age or height. The holographic face recognition system is similar to the human eye, which is equipped with a 3D visual system to acquisite the three-dimentional model of the face. The geometric accuracy that can be distinguished can reach 1/5 mm, and even the identical twins can be distinguished. Holographic face recognition system is one of biometric recognition system with the most accuracy in the market at present.


   The holographic face recognition system makes use of special holographic scanning technology and the method of structured light, which can quickly and accurately acquire the 3D model of object surface(depth)without being affected by ambient light with more accurate identification. At the same time, the textured camera and the three-dimensional sensor with wide field of view together are capable of capturing the shape, material and color of the surface of the object synchronously. With this system, the three-dimensional model of the object can be acquired in snapshot mode or video mode.


   WiMi will provide a complete solution and SDK for each field, and a certain amount of money will be charged for technical service and after-sale service. The charging mode is divided into several ways, such as charging according to the project, charging according to the number of usage, charging on annual base, the technological charges for each usage of equipment, the charge for the number of video paths in monitoring aspect and so on. It can be widely used in plenty of system management which needs accurate safety recognition, such as access control system, furniture, business office, prison, airport, customs wharf, railway station, bus station, entrance to examination room, notary office, vehicle management office, various key laboratories, key office areas, banks and securities.

1 拷贝.jpg

Product Introduction:

  1. Based on a single selfie of the user, a high-quality 3D facial model can be created.

  2. The head model supports blend shape facial animation during the operation(45 different expressions).

  3. Hairstyle is allowed to be changed, and the customized hairstyles has the capability of recoloring (more optional).

  4.The head can be added to any body with correct save. It has cartoon style avatar with skeletal animation and different rendering modes.

  5. It supports the user-generated models to be integrated into applications based on Unity, IOS,  Android and so on.

   WiMi will continue to create maximum value for customers and society in a faster, more accurate, more reliable and more convenient way.


   According to the development status of face recognition industry in prospective industry, the global biometric market will be about $12.713 billion, of which face recognition is about $2.653 billion, which accounts for about 20 percent of the all by 2016. The global face recognition market is expected to reach $6.37 billion by 2021 with a compound growth rate of 17.83 percent according to the expected period According to the scale of face recognition market in China and the present situation of face recognition industry, it is estimated that the size of face recognition market in China accounts for about 10% of the global market. From 2010 to 2016, the face recognition market in China has been growing year by year with an average annual compound growth rate of 27%. In 2016, the face recognition industry in China has a market scale of about 17.25 billion yuan with a year-on-year growth of 27.97% and a rise of 4.64 percentage points from the previous year. With the inelastic demand of online identity authentication and the cultivation of user habits in China, the application scene of face recognition will be further enlarged. Considering the security upgrading, online identity authentication and innovative Internet applications, the future market size is expected to exceed the expected growth rate.


   In addition, with the development of holographic capture and measurement technology, holographic face recognition algorithm can make up for the loss of effective recognition information caused by 2D projection. It has a good solution to the traditional difficulties, such as face rotation, occlusion, extreme similarity and so on, and has gradually become another important development route of face recognition technology. In the face recognition industry, the target customers of the enterprise mainly include the government, the enterprise and the individual. Among them, the main products for the government include security products, e-government products, counter-terrorism products and so on; the products for the enterprise mainly includes technical solutions, terminal products and so forth;etc; the products for individual customers are mainly social contact and mobile-related products. With the improvement of holographic face database and the reduction of equipment cost, holographic face technology will be more used in various industries, which has a huge market space.


   At present, there are still several core problems in the field of face recognition in the industry, while the use of the holographic face recognition system will make great improvement and provide solutions for it.


   Lighting problem: In the practical application of traditional face recognition system, the detected face image will have different shadow changes due to the change of ambient light. The error rate is large and the recognition success rate is low.


   The application of the holographic recognition method: The 3D structured light face recognition system is based on the structure light algorithm with deep learning and the deep camera with the 3D structured light, all of which is the independent research and development by WiMi. The fault tolerance rate can reach 1/1000000.


   Posture problem: The posture problem is also a technical difficulty of the present face recognition technology. At present, most of the face recognition algorithms mainly focus on the front face and quasi face images. When it happens that the face is in pitching or the deviation of the left and right side is large, the recognition rate of face recognition algorithm will also decrease sharply.


   The application of holographic recognition method: It makes use of the structured light method, which is capable of acquiring the three-dimensional model of object surface (deep)  quickly and accurately. It can capture face information from a greater angle.


   Occlusion problem: Occlusion problem is a very serious problem for face image acquisition in the condition of refusing cooperation. In the condition of wearing glasses, hat and other ornaments, It makes the face image acquisition may be incomplete, which affects the extraction and recognition of the following features, and even leads to the failure of the face detection algorithm.


   The application of the holographic recognition method: When the structured light projects specific light information to the surface of the object, it is collected by the camera. According to the change of optical signal caused by the object, the position and depth of the object are calculated, and then the whole three-dimensional space is restored. Even if wearing glasses and sunglasses, it still can be quickly recognized.


   Age change: With the change of the age, facial appearance also changes, especially for adolescents, whose change is more obvious. For different age groups, the recognition rate of face recognition algorithm is also different


   The application of the holographic recognition method: Through machine learning algorithm, computer system will learn the characteristics of the details according to the first photo for each face recognition, and more data and information will be collected each time face swiping afterwards.


   Diversified expressions: Different expressions cause different facial changes, and additionally, the effects of the same expression made by different people are also different, so it is difficult to accurately distinguish the effects of various expressions on different people by uniform criteria.


   The application of the holographic recognition method: Through the deep learning algorithm, the posture of each face swiping can not be exactly the same. The deep learning system will collect more expression data according to the characteristics of the learning details of each of your photos.


   The complex background: When the background of the monitoring scene is more complex, the face detection rate will be reduced, so the face detection algorithm which can adapt to the complex background environment is a difficult problem to be solved by face recognition technology.


   The application of the holographic recognition method: After the structured light is projected to the face, the rest of the background will not be captured, but only the three-dimensional model information of the face will be captured and compared.


Advantages of holographic face recognition:



   High recognition accuracy: It is capable of distinguishing twins and distinguishing the face with sunglasses and hats. It can work at night when the light is insufficient, and the face still can be identified while walking.


   Fast recognition speed:The registration time is 2 seconds; the identification time is less than 1 second; the database of the registrants can be up to more than 100,000; the throughout capacity is up to 60 people per minute.


   Reliability and security: No error caused by the "human factor" can deceive a high-quality 3D measured face with an ordinary photo.


   Practicability: The working distance range is 0.8 ≤ 1.6 meters, no need for direct physical contact and positioning, and which can be integrated with other equipment conveniently and simplely.


WiMi will develop such services for holographic face recognition as follows:

   High performance heterogeneous distributed cloud platform: The distributed deep learning platform supports the customized extension of deep learning model and algorithm, and support a large number of general CPU, GPU or CPU, GPU hybrid distributed computation.


   Large-scale training system with deep learning: The large-scale training system with deep learning is independently researched and developed by WiMi, which supports multi-machine and multi-GPU distributed model training with deep learning, and hundreds of billions of parameter models as well as hundreds of millions of categories of large-scale classification. The memory optimization and communication optimization technologies are in the leading role in the industry. The hundreds of GPU joint training has greatly improved the speed of company training and iterative models.


   Heterogeneous high-performance super computing platform: The computing capacity is an important force to drive the great mass fervour of current artificial intelligence. High-performance heterogeneous computing platform has more than 6000 high-performance GPUs and multiple computing clusters with central unified storage and lightweight virtualization, which provides continuous computing power support to the company's researchers.


   The heterogeneous basic algorithm library with the high performance includes a variety of machine learning algorithms, mathematics and image processing algorithms as well as deep neural network. Compared with the open source platform library in the industry, it brings 2 to 5 times improvement in performance. It supports mainstream cloud, personal computer, IOS & Android and hardware platforms with embedded port. It supports a variety of system platforms, such as Linux,Android,IOS, windows and so like.