Google Tech Talks October, 25 2007 – Speaker: John R Smith, IBM Research
In this interview, John describes current issues facing video search and the overall current state of video search whereby finding relevant video content is somewhat difficult at the moment. Reasons for this center around the fact that there is no standard as far as classification techniques. Some engines rely solely on metadata while others are utilizing recognition techniques, like(relies almost exclusively on speech recognition). John talks about the need for a true classification method centered around automatic machine learning of the content within a video itself using multiple techniques. In addition, John discusses the roll of MPEG-7 format as an ideal standard for storage of metadata.