This paper describes our participation in the TREC Video Retrieval evaluation. Our approach uses two complementary automatic approaches (the first based on visual content, the other on transcripts), to be refined in an interactive setting. The experiments focused on revealing relationships between (1) different modalities, (2) the amount of human processing, and (3) the quality of the results.
|Name||NIST Special Publication|
|Publisher||National Institute of Science en Technology (NIST)|
|Conference||Tenth Text REtrieval Conference, TREC-10 2001|
|Period||13/11/01 → 16/11/01|
- DB-MMR: MULTIMEDIA RETRIEVAL