Tokyo Tech, in collaboration with Canon Inc., achieved the world's best performance in the TRECVID2012 video search workshop for the 2nd consecutive year

Tokyo Tech achieved the best performance for the 2nd consecutive year in the Semantic Indexing task of TRECVID2012 hosted by the National Institute of Standards and Technology (NIST), outperforming 15 teams comprised of top universities and research institutions from all over the world. TRECVID is an annual international workshop of video information search technology, which attracts excellent research institutes from around the world to compete with each other on various tasks. The Semantic Indexing task involves extracting features from internet video that are meaningful for the user without using any text information such as tags or meta-data. These can include: objects such as cars and chairs, scenes such as sunsets and families having an enjoyable time, or events such as wedding ceremonies and fireworks. This task is considered to be the core task of TRECVID and our first place result indicates that Tokyo Tech has the world's best video search technology. This year we upgraded the video features and utilized the context information from the preceding and succeeding video shots. The Koichi Shinoda Lab in the Computer Science Department developed this method. Tokyo Tech's supercomputer, TSUBAME, also contributed to this achievement by providing large computational resources. This project is sponsored by Canon Inc.

Nov. 12, 2012

Contact: Koichi Shinoda, shinoda@cs.titech.ac.jp