Papers › Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification

Mitigating Embedding and Class Assignment Mismatch in Unsupervised Image Classification

1 Aug 2020ECCV 2020 8archive 2025-07-28

Sungwon Han, Sungwon Park, Sungkyu Park, Sundong Kim, Meeyoung Cha

Unsupervised image classification is a challenging computer vision task. Deep learning-based algorithms have achieved superb results, where the latest approach adopts unified losses from embedding and class assignment processes. Since these processes inherently have different goals, jointly optimizing them may lead to a suboptimal solution. To address this limitation, we propose a novel two-stage algorithm in which an embedding module for pretraining precedes a refining module that concurrently performs embedding and class assignment. Our model outperforms SOTA when tested with multiple datasets, by substantially high accuracy of 81.0% for the CIFAR-10 dataset (i.e., increased by 19.3 percent points), 35.3% accuracy for CIFAR-100-20 (9.6 pp) and 66.5% accuracy for STL-10 (6.9 pp) in unsupervised tasks.

Paper PDF Code

Code

dscig/TwoStageUC officialmentioned in paperpytorch 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.

Code Syntology ran Syntology

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

Tasks

Classification General Classification Image Classification Image Clustering Unsupervised Image Classification image-classification

Results from the paper archive 2025-07-28

Task	Dataset	Model	Metric	Value	Rank at snapshot	Leaderboard	Report
Image Clustering	CIFAR-10	TSUC	ARI	-	#28 of 40	Archive leaderboard	report
Image Clustering	CIFAR-10	TSUC	Accuracy	0.81	#28 of 40	Archive leaderboard	report
Image Clustering	CIFAR-10	TSUC	Backbone	ResNet-18	#28 of 40	Archive leaderboard	report
Image Clustering	CIFAR-10	TSUC	NMI	-	#28 of 40	Archive leaderboard	report
Image Clustering	CIFAR-10	TSUC	Train set	Train	#28 of 40	Archive leaderboard	report
Image Clustering	CIFAR-100	TSUC	Accuracy	0.353	#22 of 30	Archive leaderboard	report
Image Clustering	STL-10	TSUC	Accuracy	0.665	#21 of 29	Archive leaderboard	report
Image Clustering	STL-10	TSUC	Backbone	ResNet-18	#21 of 29	Archive leaderboard	report
Unsupervised Image Classification	CIFAR-10	TSUC	Accuracy	81.0	#7 of 9	Archive leaderboard	report
Unsupervised Image Classification	CIFAR-20	TSUC	Accuracy	35.3	#13 of 14	Archive leaderboard	report
Unsupervised Image Classification	STL-10	TSUC	Accuracy	66.50	#7 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.

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