"There is a blurry picture in the user's handset, which can be enhanced with a deep learning algorithm that magnifies the very small picture, making it a clear picture and even recognizing the small and vague human face in the picture "This is the face recognition technology in cell phone album application scenarios.
Recently, the Southern reporter interviewed the Chinese Academy of Sciences Shenzhen Institute of Advanced Technology, vice president, Professor of Chinese University of Hong Kong Tang Xiao Ou, learned that he and "face recognition" has a bond.
It is worth mentioning that in 2014, the depth face recognition technology developed by Tang Xiao-Ou led the team to achieve a recognition rate of 99.55% on the famous test benchmark LFW, which is more accurate than the human eye recognition (97.53%) and more than Facebook's D eepFace algorithm (97.35%).
"Addiction" son
Deep Learning lets machines see the world
"We always wanted to use computer vision and artificial intelligence in our daily life." Tang Xiao-Ou said he had read a large number of picture books as a child and became interested in the images. In the opinion of Tang Xiao-Ou, this may be the seed of his computer vision research.
In 2003, Tang Xiaogu's son was born, Tang Xiao Ou always back and forth in Beijing and Hong Kong, each time I see my son will take a lot of photos, and continue to shoot. Due to tens of thousands of photos, can not be managed. Later, Tang Xiao-Ou led the students to do a photo tagging project with a photo tagging system, which detected the face in the album and then used the face recognition and UI design to help auto-tagging.
In 2011, the team led by Tang Xiaogu began to study "deep learning", one of the first Chinese teams to engage in deep learning in academia. Deep learning is to dig in a large amount of data to summarize some of the data in the law, and ultimately allow the machine to learn a certain skills, such as Go, identify faces, autonomous vehicles and so on.
Tang Xiao Ou said that let the machine can understand this world is a very difficult thing.
"We looked at a computer for a large number of face shots, which might not be as strong at first, might identify many wrong people, but you mistakenly told you every time it went wrong, and it adjusted parameters based on the error. Very large number of photos, adjusted a lot of parameters, it becomes very smart. "Team members, CAS Shenzhen Institute of Advanced Technology Institute of Qiao Yu, deputy director of the South told reporters.
In 2014, face recognition achieved breakthrough results. In March of that year, Tang Xiao-Ou team published G aussianFace face recognition algorithm with an accuracy rate of 98.52%. The world's first breakthrough in human eye recognition capability (97.53%) was over the same period as the DeepFace algorithm released by Facebook in the same period (97.35%). In June of the same year, Tang Xiao-Ou's team also published the DeepID2 algorithm, raising the accuracy of face recognition to 99.15%. Then, DeepID3 algorithm in turn raised the accuracy of face recognition to 99.55%, leading the world. DeepID represents the true maturity of face recognition technology, opening up the entire face recognition industry technology.
Addiction
In security, finance and other fields have been large-scale application
At the crossroads, elevator, access and other key locations to deploy a camera to capture the face of all the people after the capture, and the "blacklist" database to do the comparison, if the comparison is successful, the system will automatically alarm. Face recognition technology began to walk into everyday life.
"Last year, China Mobile made 300 million certifications with our algorithm." Tang Xiao-Ou said face recognition has been used on a large scale in areas such as security and finance.
"Face recognition is widely used. Now many girls like beauty camera, in fact, also involves face recognition, like F aceu in the eyes with a pair of glasses, some special effects are involved face recognition." Qiao Yu told reporters Southern , Deep learning and face recognition will be used in robots, "like monitoring robots, patrol robots to find the face first, then identify whether this district, is not the family, the robot can determine whether there is any abnormal cell Case."
Tang Xiao-Ou said that after the deep learning algorithm made a breakthrough, artificial intelligence has entered a rapid development stage, began to be used in many specific industries. In face recognition, image classification, automatic driving, medical technology, and several other aspects have made many breakthroughs. But at the same time, he also pointed out that the road ahead is still long. We did not lose at the starting line this time. However, if we want to maintain the top-tier team in the world, we must have continuous investment in basic research and development and efficient linkage of production, study and research.
Last year, NVIDIA selected the Top 10 Pioneer Labs for Global Artificial Intelligence including M IT and Stanford. The laboratories of Tang Xiao-Ou are among the only selected laboratories in Asia.
It seems Tang Xiaogu, China not only research and development funding, corporate strength is not a problem, but there are a large number of overseas talent back. He believes that artificial intelligence is a new opportunity for China. "China should step up its accumulation of original technologies and make innovations at source," said Tang Xiao-Ou.
Chiller System,HVAC Chillers,Chiller Units,Industrial Water Chillers
Kassel Machinery (zhejiang) Co., Ltd. , https://www.kassel-group.com