LSD-C: linearly separable deep clusters

We present LSD-C, a novel method to identify clusters in an unlabeled dataset. Our algorithm first establishes pairwise connections in the feature space between the samples of the minibatch based on a similarity metric. Then it regroups in clusters the connected samples and enforces a linear separat...

Description complète

Détails bibliographiques
Auteurs principaux: Rebuffi, SA, Ehrhardt, S, Han, K, Vedaldi, A, Zisserman, A
Format: Conference item
Langue:English
Publié: IEEE 2021