Event Location: NSH 1507
Bio: Chen Sun is a Ph.D. candidate in the Computer Vision group at University of Southern California, advised by Prof. Ram Nevatia. His research interest includes Computer Vision and Machine Learning, with a focus on large-scale video understanding. Chen got his bachelor degree in Computer Science at Tsinghua University, Beijing. He has collaborated with researchers at Google Research and Facebook AI Research over the summers.

Abstract: The ever-increasing popularity of video capturing devices and video sharing websites creates great opportunities for researchers to utilize the rich information encoded by consumer videos. Yet understanding videos on a large scale remains challenging: the video qualities usually vary in resolution, lighting condition and camera movement; spatiotemporal annotation of the videos could be expensive and time-consuming. As videos can be naturally represented by a hierarchy of events, activities and objects, it is essential to build a pool of mid-level concepts for semantic video interpretation. In light of these challenges, my research towards video understanding focuses on utilizing weak video-level annotations effectively, and building a stronger connection between language and vision. In this talk, I will show how Internet images can be used to localize fine-grained actions in videos without using temporal video annotations. I will then describe my approach to connect language and vision via visual concepts, and demonstrate how to automatically discover the visual concepts from parallel text and visual corpora.