• Aucun résultat trouvé

IRIM at TRECVID 2013: Semantic Indexing and Instance Search

N/A
N/A
Protected

Academic year: 2021

Partager "IRIM at TRECVID 2013: Semantic Indexing and Instance Search"

Copied!
12
0
0

Texte intégral

Références

Documents relatifs

Hierarchical fusion with multiple descriptor variants and multiple classifier variants was used and optimized for the semantic indexing task.. We made several ex- periment in order

Hierarchical fusion with multiple descriptor variants and multiple classifier variants was used and optimized for the semantic indexing task.. We made several ex- periment in order

Several such diagrams are activated, using various word embeddings, image analysis models, and weights for the average, so that for given a topic and keyframe, we have

Considering only the results in table 7 (2013, 2014 and 2015 test sets), with and without I-frames, all of en- gineered features, temporal re-scoring and conceptual feed back do

The remainder of this paper briefly describes the descriptors that we have been using, the training and the various fusion schemes, and the content of the submitted runs; and the

Then, results of fusion of individual methods obtained for key frames extracted at 1fps for SIFT and 1 key frame per shot for Opponent SIFT (f5 to f 8).. At last, we display results

• M A IRIM1 1: the best of the 5 fusion approaches for each concept, followed by temporal re-scoring, conceptual feedback and a second temporal re- scoring;.. • M A IRIM2 2: similar

Our system proposes a simple scheme that combines two person recognition methods and two location recog- nition methods, performs a late fusion on each type individually, and applies