Distance-Based Image Classification: Generalizing to new classes at near-zero cost

Thomas Mensink 1 Jakob Verbeek 2 Florent Perronnin 3 Gabriela Csurka 3
2 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : We study large-scale image classification methods that can incorporate new classes and training images continuously over time at negligible cost. To this end we consider two distance-based classifiers, the k-nearest neighbor (k-NN) and nearest class mean (NCM) classifiers, and introduce a new metric learning approach for the latter. We also introduce an extension of the NCM classifier to allow for richer class representations. Experiments on the ImageNet 2010 challenge dataset, which contains over 106 training images of 1,000 classes, show that, surprisingly, the NCM classifier compares favorably to the more flexible k-NN classifier. Moreover, the NCM performance is comparable to that of linear SVMs which obtain current state-of-the-art performance. Experimentally we study the generalization performance to classes that were not used to learn the metrics. Using a metric learned on 1,000 classes, we show results for the ImageNet-10K dataset which contains 10,000 classes, and obtain performance that is competitive with the current state-of-the-art, while being orders of magnitude faster. Furthermore, we show how a zero-shot class prior based on the ImageNet hierarchy can improve performance when few training images are available.
Type de document :
Article dans une revue
IEEE Transactions on Pattern Analysis and Machine Intelligence, Institute of Electrical and Electronics Engineers, 2013, 35 (11), pp.2624-2637. 〈10.1109/TPAMI.2013.83〉
Liste complète des métadonnées

Littérature citée [48 références]  Voir  Masquer  Télécharger


https://hal.inria.fr/hal-00817211
Contributeur : Thoth Team <>
Soumis le : mercredi 24 avril 2013 - 09:59:44
Dernière modification le : mercredi 11 avril 2018 - 01:58:42
Document(s) archivé(s) le : lundi 3 avril 2017 - 09:07:10

Fichiers

mensink13pami.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

Collections

Citation

Thomas Mensink, Jakob Verbeek, Florent Perronnin, Gabriela Csurka. Distance-Based Image Classification: Generalizing to new classes at near-zero cost. IEEE Transactions on Pattern Analysis and Machine Intelligence, Institute of Electrical and Electronics Engineers, 2013, 35 (11), pp.2624-2637. 〈10.1109/TPAMI.2013.83〉. 〈hal-00817211〉

Partager

Métriques

Consultations de la notice

1222

Téléchargements de fichiers

2276