IMT Atlantique - IMT Atlantique Bretagne-Pays de la Loire (Campus Brest : Technopôle Brest-Iroise CS 8381829238 BREST Cedex 3 -
Campus Nantes : 4, rue Alfred Kastler- La chantrerie 44300 NANTES -
Campus Rennes : 2 Rue de la Châtaigneraie, 35510 CESSON SEVIGNE - France)
Abstract : In this work we present a novel unsupervised framework for hard training example mining. The only input to the method is a collection of images relevant to the target application and a meaningful initial representation, provided e.g. by pre-trained CNN. Positive examples are distant points on a single manifold, while negative examples are nearby points on different manifolds. Both types of examples are revealed by disagreements between Euclidean and manifold similarities. The discovered examples can be used in training with any discriminative loss. The method is applied to unsupervised fine-tuning of pre-trained networks for fine-grained classification and particular object retrieval. Our models are on par or are outperforming prior models that are fully or partially supervised.
https://hal.inria.fr/hal-01843085
Contributor : Yannis Avrithis
<>
Submitted on : Wednesday, July 18, 2018 - 2:46:47 PM
Last modification on : Thursday, November 15, 2018 - 11:59:01 AM
Document(s) archivé(s) le : Friday, October 19, 2018 - 7:23:06 PM
Ahmet Iscen, Giorgos Tolias, Yannis Avrithis, Ondřej Chum. Mining on Manifolds: Metric Learning without Labels. CVPR 2018 - IEEE Computer Vision and Pattern Recognition Conference, Jun 2018, Salt Lake City, United States. pp.1-10. ⟨hal-01843085⟩
