Jinlin Guo - Transfer Talk - 28th March 2013

Video Category: 
Transfer Talk
Jinlin Guo

Title: Generating Textual Description for User-generated Video Using Audio-Visual Semantic Concepts

Abstract: Due to the increasing affordability for capturing and storing video and the proliferation of Web 2.0 applications, the amount of User-generated Videos (UGVs) on the web has been skyrocketing. With regard to UGVs, metadata information is limited or not always available, which hinders many UGV-based applications such as video search. On the other hand, research in automatic detection of semantic concepts has now reached the point where hundreds of concept detectors can be built, and the performance is continuously improving, and these now offer novel opportunities for video retrieval. In such scenarios, we explore the techniques of automatically generating textual descriptions for recounting UGVs of complex events by audio-visual semantic concepts in this work.