12月28日：周吉喆

学术报告

12月28日：周吉喆

发布时间：2018-12-20 浏览量：2429

报告题目： Mosaic Streaming Videos for Privacy Protection
报告人：周吉喆澳门大学
主持人：李钦副教授
报告时间：12月28日周四14:00-15:00
报告地点：理科大楼B1002

报告摘要：
Admit it or not, live stream platforms and streamers become part of our daily life. Benefit from the popularity of smart-phones and 4G mobile networks, online live video stream and webcast are amazingly pervasive today. The content of such streams varies, including video games, talent shows, live entertainment, sports, outdoor activities, etc. Although the indoor broadcasting solely involves with the streamer alone at the most time, outdoor live streaming is the other important part. Young people holding a selfie stick, talking to their smart-phones while walking in public is a common phenomenon these days. Pushing fresh travel stories, sharing recent personal news, introducing cuisine or genuine attractions to their followers are the main contents of such outdoor live streams. During such streaming, many other people (not limited to passers and pedestrians) are filmed by camera unconsciously and then broadcast to potentially anyone without their permission. In such cases, streaming platforms and apps including Facebook and Twitch has raised a host of legal issues, primarily in the areas of privacy infringement. Different laws of privacy are applied in different regions, as a rule of thumb, respect privacy rights are universally correct. With this issue being valued by various countries, the safest approach is merely to avoid streaming any personal privacy sensitive pictures to the public.Referring to the practice of TV shows, people who are unwilling to appear in front of the camera will be mosaiced to protect their privacy. However, the placement of mosaics is completed nearly through bare hands even today. Involving with too much human labor, the manual mosaic method becomes unrealistic when we switch to massive anytime-anywhere-possible live video streaming. Thus, in this lecture, we dedicate to demonstrate our newly developed automatic filtering mechanism for personal privacy. To release the filtering burden, this new tool called MSVPP (mosaic streaming videos for privacy protection) is developed to automate such mosaic processes during live streaming. Through a series of deep CNN and learning based inferences, we can allocate proper mosaics to block people's privacy from leaking and display as much other insensitive information as possible.

报告人简介：
周吉喆，澳门大学博士，研究方向包括 Computer Vision through Deep Learning, Video Recognition & Adversal Learning, Control Theory of Complex System, Robotics and Swarm Intelligence.