Papers › SCAN: Learning to Classify Images without Labels

SCAN: Learning to Classify Images without Labels

25 May 2020ECCV 2020 8arXiv:2005.12320archive 2025-07-28

Wouter Van Gansbeke, Simon Vandenhende, Stamatios Georgoulis, Marc Proesmans, Luc van Gool

Can we automatically group images into semantically meaningful clusters when ground-truth annotations are absent? The task of unsupervised image classification remains an important, and open challenge in computer vision. Several recent approaches have tried to tackle this problem in an end-to-end fashion. In this paper, we deviate from recent works, and advocate a two-step approach where feature learning and clustering are decoupled. First, a self-supervised task from representation learning is employed to obtain semantically meaningful features. Second, we use the obtained features as a prior in a learnable clustering approach. In doing so, we remove the ability for cluster learning to depend on low-level features, which is present in current end-to-end learning approaches. Experimental evaluation shows that we outperform state-of-the-art methods by large margins, in particular +26.6% on CIFAR10, +25.0% on CIFAR100-20 and +21.3% on STL10 in terms of classification accuracy. Furthermore, our method is the first to perform well on a large-scale dataset for image classification. In particular, we obtain promising results on ImageNet, and outperform several semi-supervised learning methods in the low-data regime without the use of any ground-truth annotations. The code is made publicly available at https://github.com/wvangansbeke/Unsupervised-Classification.

PaperPDFConference PDFCode

In Syntology View this paper on Syntology, its page in Syntology's graph. That page lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

wvangansbeke/Unsupervised-Classification officialmentioned in papermentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to A Simple Framework for Contrastive Learning of Visual Representations (arXiv:2002.05709; that paper's own run record: 115 ran (of which 33 constructed an object rather than computing a result; 89 with no instrument failure: 1 honoured, 1 violated, 87 with no contract checked; 26 where Syntology's instrument failed) · 22 unverified); nothing checks that they implement this paper's method. report
2023-MindSpore-4/Code14 mindsporeSyntology: not harvested report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

ClassificationClusteringGeneral ClassificationImage ClassificationImage ClusteringRepresentation LearningSemi-Supervised Image ClassificationUnsupervised Image Classificationimage-classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Clustering CIFAR-10 SCAN ARI 0.772 #18 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN Accuracy 0.883 #18 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN Backbone ResNet-18 #18 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN NMI 0.797 #18 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN Train set Train #18 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN (Avg) ARI 0.758 #19 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN (Avg) Accuracy 0.876 #19 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN (Avg) Backbone ResNet-18 #19 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN (Avg) NMI 0.787 #19 of 40 Archive leaderboard report
Image Clustering CIFAR-10 SCAN (Avg) Train set Train #19 of 40 Archive leaderboard report
Image Clustering CIFAR-100 SCAN ARI 0.333 #13 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN Accuracy 0.507 #13 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN NMI 0.486 #13 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN Train Set Train #13 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN (Avg) ARI 0.301 #16 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN (Avg) Accuracy 0.459 #16 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN (Avg) NMI 0.468 #16 of 30 Archive leaderboard report
Image Clustering CIFAR-100 SCAN (Avg) Train Set Train #16 of 30 Archive leaderboard report
Image Clustering ImageNet SCAN Accuracy 39.9 #11 of 12 Archive leaderboard report
Image Clustering ImageNet SCAN NMI 72.0 #11 of 12 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) SCAN ACCURACY 0.662 #5 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) SCAN ARI 0.544 #5 of 5 Archive leaderboard report
Image Clustering ImageNet-100 (TEMI Split) SCAN NMI 0.787 #5 of 5 Archive leaderboard report
Image Clustering ImageNet-200 SCAN ACCURACY 0.563 #4 of 5 Archive leaderboard report
Image Clustering ImageNet-200 SCAN ARI 0.441 #4 of 5 Archive leaderboard report
Image Clustering ImageNet-200 SCAN NMI 0.757 #4 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) SCAN ACCURACY 0.751 #5 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) SCAN ARI 0.635 #5 of 5 Archive leaderboard report
Image Clustering ImageNet-50 (TEMI Split) SCAN NMI 0.805 #5 of 5 Archive leaderboard report
Image Clustering STL-10 SCAN Accuracy 0.809 #15 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN Backbone ResNet-18 #15 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN NMI 0.698 #15 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN Train Split Train #15 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN (Avg) Accuracy 0.767 #16 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN (Avg) Backbone ResNet-18 #16 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN (Avg) NMI 0.680 #16 of 29 Archive leaderboard report
Image Clustering STL-10 SCAN (Avg) Train Split Train #16 of 29 Archive leaderboard report
Semi-Supervised Image Classification ImageNet - 1% labeled data SCAN (ResNet-50|Unsupervised) Top 1 Accuracy 39.90% #53 of 65 Archive leaderboard report
Semi-Supervised Image Classification ImageNet - 1% labeled data SCAN (ResNet-50|Unsupervised) Top 5 Accuracy 60.0% #53 of 65 Archive leaderboard report
Unsupervised Image Classification CIFAR-10 SCAN Accuracy 88.3 #5 of 9 Archive leaderboard report
Unsupervised Image Classification CIFAR-20 SCAN Accuracy 50.7 #10 of 14 Archive leaderboard report
Unsupervised Image Classification ImageNet SCAN (ResNet-50) ARI 27.5 #7 of 9 Archive leaderboard report
Unsupervised Image Classification ImageNet SCAN (ResNet-50) Accuracy (%) 39.9 #7 of 9 Archive leaderboard report
Unsupervised Image Classification STL-10 SCAN Accuracy 80.90 #6 of 9 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Introduced by this paper: SCAN-clustering

1x1 ConvolutionAverage PoolingBatch NormalizationBottleneck Residual BlockColorJitterConvolutionDense ConnectionsFeedforward NetworkGlobal Average PoolingInfoNCEKaiming InitializationMax PoolingMoCo v2NT-XentRandom Gaussian BlurRandom Resized CropReLUResidual BlockResidual ConnectionSCAN-clusteringSimCLRk-Means Clustering

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections