Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Joint Representation Learning and Novel Category Discovery on Single- and Multi-modal Data

About

This paper studies the problem of novel category discovery on single- and multi-modal data with labels from different but relevant categories. We present a generic, end-to-end framework to jointly learn a reliable representation and assign clusters to unlabelled data. To avoid over-fitting the learnt embedding to labelled data, we take inspiration from self-supervised representation learning by noise-contrastive estimation and extend it to jointly handle labelled and unlabelled data. In particular, we propose using category discrimination on labelled data and cross-modal discrimination on multi-modal data to augment instance discrimination used in conventional contrastive learning approaches. We further employ Winner-Take-All (WTA) hashing algorithm on the shared representation space to generate pairwise pseudo labels for unlabelled data to better predict cluster assignments. We thoroughly evaluate our framework on large-scale multi-modal video benchmarks Kinetics-400 and VGG-Sound, and image benchmarks CIFAR10, CIFAR100 and ImageNet, obtaining state-of-the-art results.

Xuhui Jia, Kai Han, Yukun Zhu, Bradley Green• 2021

Related benchmarks

TaskDatasetResultRank
Generalized Category DiscoveryImageNet-100
All Accuracy33.1
208
Generalized Category DiscoveryCIFAR-100
Accuracy (All)44.1
185
Category DiscoveryStanford Cars
Accuracy (All)20
71
Category DiscoveryCUB-200 2011
Overall Score26.5
71
Generalized Category DiscoveryCUB-200 (test)
Overall Accuracy26.5
63
Category DiscoveryCIFAR10
Accuracy (All)65.4
60
Generalized Category DiscoveryOxford Pets
Accuracy (All)35.2
50
Category DiscoveryFood101
Accuracy (All)18.2
45
Fine-grained object category discoveryStanford Cars (test)--
38
Category DiscoveryCIFAR-100
Accuracy (All Categories)44.1
18
Showing 10 of 23 rows

Other info

Follow for update