A Comparison of User Strategies in Image Retrieval with Relevance Feedback
Abstract
Given the difficulty of setting up large-scale experiments with real users, the comparison of content-based image retrieval methods using relevance feedback usually relies on an emulation of the user, following a single, well-prescribed strategy. Since the behavior of real users cannot be expected to comply to strict specifications, it is very important to evaluate the sensitiveness of the retrieval results to likely variations of users' behavior. It is also important to find out whether some strategies help the system to perform consistently better, so as to promote their use. We compare here two algorithms of SVM-based relevance feedback using the angular kernel. In these experiments, the user is emulated according to seven significantly different strategies on four ground-truth databases of different complexity. We first find that the ranking of the two algorithms does not depend much on the selected strategy. Second, the ranking between strategies appears to be relatively independent of the complexity of the ground-truth classes, which allows us to identify desirable characteristics in the behavior of the user.
Reference
@inproceedings{jpt-delos05,
author = {Crucianu, M. and Tarel, J.-P. and Ferecatu, M.},
title = {A Comparison of User Strategies in Image Retrieval with Relevance Feedback},
booktitle = {Proceedings of 7th International Workshop on Audio-Visual Content and
Information Visualization in Digital Libraries (AVIVDiLib'05)},
date = {May 4-6},
address = {Cortona, Italy},
pages = {121 - 130},
year = {2005},
note = {http://perso.lcpc.fr/tarel.jean-philippe/publis/delos05.html}
}
Pdf file (126 Kb)
(c) DELOS