In the field of face recognition, domestic research teams have already taken the lead in the world. Recently, Tencent U-Lab has won the top spot in the 1 million face recognition test (Challenge1/FaceScrub idenTIficaTIon) with the latest achievement of 83.290% in MegaFace, the most authoritative and difficult face recognition database in the world. The game also includes Google, Baidu and other teams. This is also the case after the recent LFW, Uto Labs once again set a world record.
About MegaFace: MegaFace is a set of public face datasets published and maintained by the University of Washington Computer Science and Engineering Laboratory. The LFW database focuses on comparing two face photos with the same identity, MegaFace With the recognition rate of massive face registration, MegaFace is currently the most authoritative and popular evaluation face recognition performance, especially one of the benchmark references for massive face recognition retrieval performance.
Conquering the massive face recognition puzzle and winning the international competition
In the field of artificial intelligence, the application scale of face recognition technology is expanding and the algorithm capability is constantly improving, which makes it difficult to distinguish the gap between different algorithm capabilities by simple face comparison evaluation. In the massive face recognition test, the face recognition accuracy rate when the interference set reaches 1 million people has become the most effective indicator for face recognition technology in finding and chasing. Tencent's team has overcome the problem of massive face recognition, and refreshing the MegaFace record is more realistic.
The training data of Tencent Utu Lab comes from the database of Westerners and Orientals it collects. Among them, Westerners have a total of 20,000 people with 2 million faces, and the Orientals have a total of 50,000 faces with 4 million faces. Youtu uses the self-developed TensorFlow-based multi-machine multi-card cluster training platform RapidFlow to train the deep learning model. After integrating hundreds of deep depth learning models, Uto finally achieved a good score of 83.290% in the face recognition accuracy of MegaFace. It is worth mentioning that, using this model, the Uight face recognition algorithm simultaneously refreshes records in both LFW and MegaFace databases, which also indicates the generalization and universality of this algorithm.
Finding relatives, security, finance, etc.
After overcoming technical problems, massive face recognition technology has a strong practical significance and can be widely used in the fields of family search and tracking. Uto's massive face recognition technology has repeatedly recorded internationally, and it also provides technical advantages for its larger-scale landing applications.
In the aspect of finding relatives, Uto has played an important role in the near future. Relying on the massive face retrieval technology developed by Utopia Lab, the “Lost You†anti-lost platform jointly issued by Fujian Provincial Public Security Department and Tencent Internet + Cooperation Department has been in the province of Fujian since it was launched in March this year. There have been nearly 10 successful cases of contact with relatives of lost people within 24 hours. In addition, Uto's face recognition technology is also widely used in a variety of products, from the internal QQ space, micro-bank, etc., to the external Drip travel, Unicom; on the other hand, Uto figure Face recognition technology is applied to life, and it is gradually expanding into areas such as finance, security, and network security.
"Developing artificial intelligence, scenarios, big data, computing power and talent are all indispensable," said Ma Huateng, chairman and CEO of Tencent, at the Shenzhen IT Leaders Summit. "The success of research can only produce value" is also the direction of teamwork and efforts that Uto has always been adhering to. For the future, Youtu said that it will create more application scenarios with an open mind, so that technology can truly make meaning to the public in reality and bring more safety and convenience to human life.
2.1 Home Theater Speaker,2.1 Speaker System Home Theater,Surround Sound Speakers,2.1 Home Theater
GUANGZHOU SOWANGNY ELECTRONIC CO.,LTD , https://www.jerry-power.com