JapaneseEnglish





NII TV-RECS (NII TV Broadcast Video Research Corpus) is an experimental prototype of a research-purpose broadcast video corpus, which is built by a group of researchers in NII aiming at an effective resource for video processing research.  Generally, a corpus is a comprehensive collection of research material.  Video corpora are expected to be useful for video processing research, in the same way as text corpora (e.g., language corpus, web corpus, news corpus, etc.) are useful for natural-language processing research.
Based on this viewpoint, we have constructed a broadcast video archiving system which has sufficient capacity and functionality to serve as a testbed for video processing research.  In order to reflect the important nature of the huge broadcast video corpus, the system captures multiple channels simultaneously, 24 hours a day, as MPEG video files.  The system also captures related text information including closed-caption text and electronic program guide (EPG).
The archiving system has 10TB storage, which is split into 5TB for 24-hour broadcast streams and 5TB for particular programs.  Due to the disk capacity limitation, the former 5TB storage keeps video files of the recent 3-weeks (about 3,500 hours in total) in MPEG-1 format.  On the other hand, the latter stores video files of particular genres of TV programs persistently, including news, documentary, cooking, etc.
Among particular programs, a news program from NHK (the largest broadcast station inJapan) has been captured almost everyday for more than seven years since March 2001.  Although capturing was not made on some days due to system maintenance, more than 2,000 days of the programs (amounting more than 1,000 hours) have been captured.  For this program, video streams are also captured in MPEG-2 format to guarantee higher quality for the purpose of high-precision video processing such as face detection and motion analysis.

In addition, by preparing over 100TB storage space at Chiba Annex, the long-term archiving of the 24-hour broadcast streams has been started since August 2009.




Publications
Publications using TV-RECS
Publications on TV-RECS
Publications related to TV-RECS



Collaborating Researchers
Name Affiliation
Shin'ichi Satoh Professor , Digital Content and Media Sciences Research Division
National Institute of Infomatics
Norio Katayama Associate Professor , Digital Content and Media Sciences Research Division
National Institute of Infomatics
Hiroshi Mo Assistant Professor , Digital Content and Media Sciences Research Division
National Institute of Infomatics
Ichiro Ide Associate Professor , Graduate School of Information Science
Nagoya University



Copyright(C) 2008-2018 National Institute of Informatics