Filters








2 Hits in 2.8 sec

CLIPS-LSR-NII Experiments at TRECVID 2005

Stéphane Ayache, Georges Quénot, Jérôme Gensel, Shin'ichi Satoh
2005 TREC Video Retrieval Evaluation  
This paper presents the systems used by CLIPS-IMAG laboratory. We participated to shot segmentation and high-level extraction tasks.  ...  For these experiments, we manually did that. Each concept's context is defined by at most 5 intermediate concepts (patchesconcept).  ...  In TRECVID'04 experiments, [9] used a basis of 22 intermediate concepts. By adding the 10 TRECVID's concepts, they learned semantic context of 32 concepts with Stacking.  ... 
dblp:conf/trecvid/AyacheQGS05 fatcat:hgrm2ym6fzf2vel7vfihoxxkg4

TRECVID 2005 - An Overview

Paul Over, Tzveta Ianeva, Wessel Kraaij, Alan F. Smeaton
2005 TREC Video Retrieval Evaluation  
9 Authors' note TRECVID would not happen without support from ARDA/DTO and NIST and the research community is very grateful for this.  ...  Richard Wright at the BBC Archive made the rushes data available and Gary Marchionini and the Open Video Project at the University of North Carolina at Chapel Hill helped us get the NASA videos in MPEG  ...  CLIPS-LSR-NII explored the use of a 3-level network of stacked classifiers based only on visual information.  ... 
dblp:conf/trecvid/OverIKS05 fatcat:czlfoelhrnaxnnhwr74antamqi