by Computer vision has been developed for more than 50 years. It can not only recognize various objects, but also distinguish the relationship between objects, and can also see details that humans cannot see. The AI Research Center of Stanford University in the United States will use it to create a computer that helps with housework. robot. Li Feifei, the director of the center and the founder of ImageNet, came to Taiwan a few days ago to reveal the new progress of the international prospective research on computer vision. From the past, present and future of computer vision, she pointed out three major technical issues, as well as the latest research progress derived from each stage. For example, in the field of object detection, the ability of object detection AI to interpret the relationship between objects has been extended from static images to dynamic videos. It can not only identify various activities in videos, but also distinguish the relationship between participating characters and objects. Moreover, object detection has also developed more advanced applications and privacy protection technologies with more fine-grained recognition, such as the human motion detection model PriHAR, which can not only accurately identify human behavior, but also directly blur human body images at the lens end. Protect personal privacy at the source.
