Because social networks and electronic imaging products are very developed now, people can easily get other people's face photos and then cheat the face recognition system, so it is very important to distinguish whether the face in front of the system is a real face or a photo/video/three-dimensional mask.
For photo cheating, it is mainly distinguished according to resolution (the resolution of remake photos is different from that of photos collected directly from real people in quality and resolution), three-dimensional information, eye movement and so on. For video deception, it is distinguished according to three-dimensional information, light and so on. There is little research on 3D mask deception (and there are relatively few actual deception scenes).
In my opinion, computer algorithms can distinguish photos from real people. But in practical application, face forensics still has great limitations. The current algorithms are basically based on laboratory data, and there is still a long way to go before they are practical.
In addition, fingerprints, irises, etc. Also face this problem.
Yun+Community 3rd Anniversary Advertisement
20 17 12, released by Cloud+Community, has a variety of community products from the initial technical blog to now. In the future, we will ride the wind and waves together to create unlimited possibilities.
Face recognition
Do I need to apply for face recognition access? You can log in to the face recognition console and click to open it immediately, without waiting for manual review. In which countries and regions can face recognition be used? At present, China can visit other countries normally, but the speed is slow. What is the access speed of face recognition? The specific response time is also related to the photo size and network environment. How to authenticate, sign or appear? ...
Face recognition access control attendance
Tencent Cloud Face Recognition Access Control Attendance Solution, based on UTU face recognition technology, solves the non-inductive demand of access gates, building visitors, attendance check-in and other scenes, and can provide partners with rapid integration of sdkapp ... face recognition and obtain recognition results, and support offline online face search to switch applicable objects. Customers have specialized software teams to develop, and customers have more business processes. Face recognition is only applicable to some functional schemes. ...
Version correlation of face recognition
What's the difference between different versions? There are currently three versions of Tencent Cloud Shentu face recognition, namely 1.0, 2.0 and 3.0. 2.0 and 3.0 are currently open versions and can be used normally. The version of 1.0 currently only supports old customers ... Can the old version of face recognition continue to be used? The maintenance of face recognition version 1.0 stopped from June 20 19. If you are still using the 1.0 version of the service, in order not to affect the normal development of your business, I suggest you do it as soon as possible. ...
Financial Cloud Face Recognition
Face verification: it can judge whether a given face picture is the same person and the similarity with someone's face. Face recognition: detect a given face picture, retrieve his identity information from the database and return it. In vivo detection: it can effectively reject the attacks of photos, videos and head models in the face verification process, and ensure that real people participate in the face verification process. Ocr identification: shoot identification documents and extract the information on them. ...