Efficient Video Annotations by an Image Groups

Abstract

Searching desirable events in uncontrolled videos is a challenging task. So, researches mainly focus on obtaining concepts from numerous labelled videos. But it is time consuming and labour expensive to collect a large amount of required labelled videos for training event models under various condition. To avoid this problem, we propose to leverage abundant Web images for videos since Web images contain a rich source of information with many events roughly annotated and taken under various conditions. However, information from the Web is difficult .so,brute force knowledge transfer of images may hurt the video annotation performance. so, we propose a novel Group-based Domain Adaptation learning framework to leverage different groups of knowledge (source target) queried from the Web image search engine to consumer videos (domain target). Different from old methods using multiple source domains of images, our method makes the Web images according to their intrinsic semantic relationships instead of source. Specifically, two different types of groups ( event-specific groups and concept specific groups) are exploited to respectively describe the event-level and concept-level semantic meanings of target-domain videos.