Our new X account is live! Follow @wizwand_team for updates
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 DiscoveryCUB-200 (test)
Overall Accuracy26.5
63
Fine-grained object category discoveryStanford Cars (test)--
38
ClusteringCIFAR10 unlabelled (train)
Clustering Accuracy93.4
14
ClusteringImageNet unlabelled (train)
Clustering Accuracy86.7
14
ClusteringCIFAR100-20 unlabelled (train)
Clustering Accuracy76.4
13
Continuous Novel Category DiscoveryCIFAR-100 DI scenario
Mf0.3
5
On-the-fly Category DiscoveryArachnida (test)
Accuracy (All)28.1
5
On-the-fly Category DiscoveryAnimalia (test)
Accuracy (All)33.4
5
On-the-fly Category DiscoveryOxford Pets (test)
Accuracy (All)35.2
5
On-the-fly Category DiscoveryFungi (test)
Accuracy (All)27.5
5
Showing 10 of 12 rows

Other info

Follow for update