Papers › Large-scale Unsupervised Semantic Segmentation

Large-scale Unsupervised Semantic Segmentation

6 Jun 2021arXiv:2106.03149archive 2025-07-28

ShangHua Gao, Zhong-Yu Li, Ming-Hsuan Yang, Ming-Ming Cheng, Junwei Han, Philip Torr

Empowered by large datasets, e.g., ImageNet, unsupervised learning on large-scale data has enabled significant advances for classification tasks. However, whether the large-scale unsupervised semantic segmentation can be achieved remains unknown. There are two major challenges: i) we need a large-scale benchmark for assessing algorithms; ii) we need to develop methods to simultaneously learn category and shape representation in an unsupervised manner. In this work, we propose a new problem of large-scale unsupervised semantic segmentation (LUSS) with a newly created benchmark dataset to help the research progress. Building on the ImageNet dataset, we propose the ImageNet-S dataset with 1.2 million training images and 50k high-quality semantic segmentation annotations for evaluation. Our benchmark has a high data diversity and a clear task objective. We also present a simple yet effective method that works surprisingly well for LUSS. In addition, we benchmark related un/weakly/fully supervised methods accordingly, identifying the challenges and possible directions of LUSS. The benchmark and source code is publicly available at https://github.com/LUSSeg.

PaperPDFCode

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

Code

LUSSeg/ImageNet-S officialmentioned on GitHubpytorch report
LUSSeg/ImageNetSegModel mentioned on GitHubpytorchNOASSERTION report
LUSSeg/PASS mentioned on GitHubpytorch 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

DiversityRepresentation LearningSegmentationSemantic SegmentationUnsupervised Semantic Segmentation

Datasets

Introduced by this paper, per the archive.

ImageNet-S

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Semantic Segmentation ImageNet-S PASS (ResNet-50 D16, 224x224, LUSS) mIoU (test) 20.8 #19 of 20 Archive leaderboard report
Semantic Segmentation ImageNet-S PASS (ResNet-50 D16, 224x224, LUSS) mIoU (val) 21.6 #19 of 20 Archive leaderboard report
Semantic Segmentation ImageNet-S PASS (ResNet-50 D32, 224x224, LUSS) mIoU (test) 20.3 #20 of 20 Archive leaderboard report
Semantic Segmentation ImageNet-S PASS (ResNet-50 D32, 224x224, LUSS) mIoU (val) 21.0 #20 of 20 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S PASS mIoU (test) 11.0 #1 of 1 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S PASS mIoU (val) 11.5 #1 of 1 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-300 PASS mIoU (test) 18.1 #1 of 1 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-300 PASS mIoU (val) 18 #1 of 1 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-50 PASS (+Saliency map) mIoU (test) 42.3 #1 of 5 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-50 PASS (+Saliency map) mIoU (val) 43.3 #1 of 5 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-50 PASS mIoU (test) 32.0 #2 of 5 Archive leaderboard report
Unsupervised Semantic Segmentation ImageNet-S-50 PASS mIoU (val) 32.4 #2 of 5 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