Papers › Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling

Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling

21 Sep 2023CVPR 2024 1arXiv:2309.12378archive 2025-07-28

Leon Sick, Dominik Engel, Pedro Hermosilla, Timo Ropinski

Traditionally, training neural networks to perform semantic segmentation required expensive human-made annotations. But more recently, advances in the field of unsupervised learning have made significant progress on this issue and towards closing the gap to supervised algorithms. To achieve this, semantic knowledge is distilled by learning to correlate randomly sampled features from images across an entire dataset. In this work, we build upon these advances by incorporating information about the structure of the scene into the training process through the use of depth information. We achieve this by (1) learning depth-feature correlation by spatially correlate the feature maps with the depth maps to induce knowledge about the structure of the scene and (2) implementing farthest-point sampling to more effectively select relevant features by utilizing 3D sampling techniques on depth information of the scene. Finally, we demonstrate the effectiveness of our technical contributions through extensive experimentation and present significant improvements in performance across multiple benchmark datasets.

PaperPDFConference PDFCode

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

Code

leonsick/depthg 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

Feature CorrelationSemantic SegmentationUnsupervised Panoptic SegmentationUnsupervised Semantic Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Unsupervised Panoptic Segmentation Cityscapes DepthG + CutLER PQ 16.1 #5 of 5 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-B/8) Clustering [Accuracy] 58.6 #7 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-B/8) Clustering [mIoU] 29.0 #7 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-B/8) Linear Classifier [Accuracy] 75.5 #7 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-B/8) Linear Classifier [mIoU] 41.6 #7 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG w/ 3D-LHP (ViT-S/8) Clustering [Accuracy] 55.1 #11 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG w/ 3D-LHP (ViT-S/8) Clustering [mIoU] 26.7 #11 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG w/ 3D-LHP (ViT-S/8) Linear Classifier [Accuracy] 73.9 #11 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG w/ 3D-LHP (ViT-S/8) Linear Classifier [mIoU] 37.8 #11 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-S/8) Clustering [Accuracy] 56.3 #14 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-S/8) Clustering [mIoU] 25.6 #14 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-S/8) Linear Classifier [Accuracy] 73.7 #14 of 29 Archive leaderboard report
Unsupervised Semantic Segmentation COCO-Stuff-27 DepthG (ViT-S/8) Linear Classifier [mIoU] 38.9 #14 of 29 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