Papers › DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation

DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation

18 Sep 2023arXiv:2309.09668archive 2025-07-28

Bowen Yin, Xuying Zhang, Zhongyu Li, Li Liu, Ming-Ming Cheng, Qibin Hou

We present DFormer, a novel RGB-D pretraining framework to learn transferable representations for RGB-D segmentation tasks. DFormer has two new key innovations: 1) Unlike previous works that encode RGB-D information with RGB pretrained backbone, we pretrain the backbone using image-depth pairs from ImageNet-1K, and hence the DFormer is endowed with the capacity to encode RGB-D representations; 2) DFormer comprises a sequence of RGB-D blocks, which are tailored for encoding both RGB and depth information through a novel building block design. DFormer avoids the mismatched encoding of the 3D geometry relationships in depth maps by RGB pretrained backbones, which widely lies in existing methods but has not been resolved. We finetune the pretrained DFormer on two popular RGB-D tasks, i.e., RGB-D semantic segmentation and RGB-D salient object detection, with a lightweight decoder head. Experimental results show that our DFormer achieves new state-of-the-art performance on these two tasks with less than half of the computational cost of the current best methods on two RGB-D semantic segmentation datasets and five RGB-D salient object detection datasets. Our code is available at: https://github.com/VCIP-RGBD/DFormer.

PaperPDFCode

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

Code

VCIP-RGBD/DFormer mentioned in papermentioned on GitHubpytorchMIT 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

3D geometryDecoderObject DetectionRGB-D Salient Object DetectionRepresentation LearningSalient Object DetectionSegmentationSemantic Segmentationobject-detection

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
RGB-D Salient Object Detection DES DFormer-L Average MAE 0.013 #1 of 13 Archive leaderboard report
RGB-D Salient Object Detection DES DFormer-L S-Measure 94.8 #1 of 13 Archive leaderboard report
RGB-D Salient Object Detection DES DFormer-L max E-Measure 98.0 #1 of 13 Archive leaderboard report
RGB-D Salient Object Detection DES DFormer-L max F-Measure 95.6 #1 of 13 Archive leaderboard report
RGB-D Salient Object Detection NJU2K DFormer-L Average MAE 0.023 #1 of 27 Archive leaderboard report
RGB-D Salient Object Detection NJU2K DFormer-L S-Measure 93.7 #1 of 27 Archive leaderboard report
RGB-D Salient Object Detection NJU2K DFormer-L max E-Measure 96.4 #1 of 27 Archive leaderboard report
RGB-D Salient Object Detection NJU2K DFormer-L max F-Measure 94.6 #1 of 27 Archive leaderboard report
RGB-D Salient Object Detection NLPR DFormer-L Average MAE 0.016 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection NLPR DFormer-L S-Measure 94.2 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection NLPR DFormer-L max E-Measure 97.1 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection NLPR DFormer-L max F-Measure 93.9 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection SIP DFormer-L Average MAE 0.032 #1 of 16 Archive leaderboard report
RGB-D Salient Object Detection SIP DFormer-L S-Measure 91.5 #1 of 16 Archive leaderboard report
RGB-D Salient Object Detection SIP DFormer-L max E-Measure 95.0 #1 of 16 Archive leaderboard report
RGB-D Salient Object Detection SIP DFormer-L max F-Measure 93.8 #1 of 16 Archive leaderboard report
RGB-D Salient Object Detection STERE DFormer-L Average MAE 0.030 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection STERE DFormer-L S-Measure 92.3 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection STERE DFormer-L max E-Measure 95.2 #1 of 14 Archive leaderboard report
RGB-D Salient Object Detection STERE DFormer-L max F-Measure 92.9 #1 of 14 Archive leaderboard report
Semantic Segmentation NYU Depth v2 DFormer-L Mean IoU 57.2% #16 of 121 Archive leaderboard report
Semantic Segmentation NYU Depth v2 DFormer-B Mean IoU 55.6% #25 of 121 Archive leaderboard report
Semantic Segmentation NYU Depth v2 DFormer-S Mean IoU 53.6% #33 of 121 Archive leaderboard report
Semantic Segmentation NYU Depth v2 DFormer-T Mean IoU 51.8% #50 of 121 Archive leaderboard report
Semantic Segmentation SUN-RGBD DFormer-L Mean IoU 52.5% #10 of 44 Archive leaderboard report
Semantic Segmentation SUN-RGBD DFormer-B Mean IoU 51.2% #15 of 44 Archive leaderboard report
Semantic Segmentation SUN-RGBD TokenFusion (S) Mean IoU 50.0% #19 of 44 Archive leaderboard report
Semantic Segmentation SUN-RGBD FSFNet Mean IoU 48.8% #27 of 44 Archive leaderboard report
Semantic Segmentation SYN-UDTIRI DFormer IoU 90.88 #7 of 10 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