Model learning for image understanding from video archives

点击翻译按钮获取中文摘要

In this study, we address techniques to automatically learn models for image understanding from everyday broadcast video. This is especially useful to realize video content analysis for video archive retrieval and to realize robot vision for interactive robot supporting daily life.For these purposes, it is not necessary to realize very precise visual measurement which is typically used for industrial robots, however, it is desired to identify many various types of objects which ordinary people know as the common sense, or to identify new object which might become new trend. The main goal of the study is to establish the methodology to automatically and dynamically obtain image understanding models from broadcast videos.