[LLC19] From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process

Conférence Internationale avec comité de lecture : 23rd International Conference on Multimedia Modeling, January 2019, Vol. 11296, pp.1-13, Series LNCS, Thessaloniki, Grece, (DOI: 978-3-030-05716-9 _ 38)
Résumé: Zero-shot learning (ZSL) is concerned with the recognition of previously unseen classes. It relies on additional semantic knowledge for which a mapping can be learned with training examples of seen classes. While classical ZSL considers the recognition performance on unseen classes only, generalized zero-shot learning (GZSL) aims at maximizing performance on both seen and unseen classes. In this paper, we propose a new process for training and evaluation in the GZSL setting; this process addresses the gap in performance between samples from unseen and seen classes by penalizing the latter, and enables to electhyper-parameters well-suited to the GZSL task. It can be applied to any existing ZSL approach and leads to a significant performance boost: the experimental evaluation shows that GZSL performance, averaged over eight state-of-the-art methods, is improved from 28.5 to 42.2 on CUB and from 28.2 to 57.1 on AwA2.

Collaboration: CEA


@inproceedings {
title="{From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process}",
author=" Y. LE CACHEUX and H. Le Borgne and M. Crucianu ",
booktitle="{23rd International Conference on Multimedia Modeling}",
address="Thessaloniki, Grece",
doi="978-3-030-05716-9 _ 38",