Savoirs en partage

Le portail des ressources scientifiques
et pédagogiques de l'AUF

LIG-MRIM at NTCIR-12 Lifelog Semantic Access Task

Sujet: Visual concepts, deep learning, temporal description, Ima- geNet, TRECVid, [INFO.INFO-IR] Computer Science [cs]/Information Retrieval [cs.IR]
Auteur: Safadi, Bahjat, Mulhem, Philippe, Quénot, Georges, Chevallet, Jean-Pierre
Résumé: This paper describes the participation of the LIG-MRIM research team to the Lifelog Semantic Access sub-task of the NTCIR-12 (2016). Our approach mainly relies on mapping the query terms to visual concepts computed on the Lifel-ogs images according to two separated learning schemes. A post-processing is then performed if the topic is related to temporal, location or activity information associated with the images. The results obtained are promising for a first participation to such a task, with event-based MAP above 29% and an event-based nDCG value close to 39%. Team Name MRIM Subtasks Lifelog Semantic Access Task
Editeur: HAL CCSD