IRIM at TRECVID 2010: Semantic Indexing and Instance Search

David Gorisse (1) , Frédéric Precioso (2, 1) , Philippe-Henri Gosselin (1) , Lionel Granjon (3) , Denis Pellerin (4) , Michèle Rombaut (4) , Hervé Bredin (5, 6) , Lionel Koenig (5) , Rémi Vieux (7) , Boris Mansencal (7) , Jenny Benois-Pineau (7) , Hugo Boujut (7) , Claire Morand (7) , Hervé Jégou (8) , Stéphane Ayache (9) , Bahjat Safadi (10) , Yubing Tong (10) , Franck Thollard (10) , Georges Quénot (10) , Matthieu Cord (2) , Alexandre Benoît (11) , Patrick Lambert (11)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
2 MALIRE - Machine Learning and Information Retrieval
3 GIPSA-Services - GIPSA-Services
4 GIPSA-AGPIG - GIPSA - Architecture, Géométrie, Perception, Images, Gestes
5 IRIT-SAMoVA - Équipe Structuration, Analyse et MOdélisation de documents Vidéo et Audio
6 CNRS - Centre National de la Recherche Scientifique
7 LaBRI - Laboratoire Bordelais de Recherche en Informatique
8 TEXMEX - Multimedia content-based indexing
9 LIF - Laboratoire d'informatique Fondamentale de Marseille - UMR 6166
10 LIG - Laboratoire d'Informatique de Grenoble
11 LISTIC - Laboratoire d'Informatique, Systèmes, Traitement de l'Information et de la Connaissance

The IRIM group is a consortium of French teams working on Multimedia Indexing and Retrieval. This paper describes our participation to the TRECVID 2010 semantic indexing and instance search tasks. For the semantic indexing task, we evaluated a number of different descriptors and tried different fusion strategies, in particular hierarchical fusion. The best IRIM run has a Mean Inferred Average Precision of 0.0442, which is above the task median performance. We found that fusion of the classification scores from different classifier types improves the performance and that even with a quite low individual performance, audio descriptors can help. For the instance search task, we used only one of the example images in our queries. The rank is nearly in the middle of the list of participants. The experiment showed that HSV features outperform the concatenation of HSV and Edge histograms or the Wavelet features.

Mots clés: descriptors; classification; high level features extraction; Semantic Indexing; Instance Search; fusion

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]; Recherche d'information [cs.IR]

David Gorisse, Frédéric Precioso, Philippe-Henri Gosselin, Lionel Granjon, Denis Pellerin, et al.. IRIM at TRECVID 2010: Semantic Indexing and Instance Search. TRECVID 2010 - TREC Video Retrieval Evaluation workshop, Nov 2010, Gaithersburg, MD, United States. ⟨hal-00591099⟩