List your company
ResearchFormerly WizWand

Semantic Segmentation on NYU-depth v2 (IoU and Pixel Accuracy)

58.8mIoURSGMamba-B
28.01636.0084451.992Jul 29, 2021May 11, 2022Feb 22, 2023Dec 6, 2023Sep 17, 2024Jul 1, 2025Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
RSGMamba-BBackbone=SegMAN-B, Params.=48.6M, Input size=480 × 640, Flops=130.2G2026.0458.8——
DFormerV2Backbone=DFormerV2-L, Params.=95.5M, Input size=480 × 640, Flops=124.1G2026.0458.4——
DFNetBackbone=MiT-B4, Params.=108.8M, Input size=480 × 640, Flops=101.0G2026.0457.9——
PrimKDBackbone=MiT-B4, Params.=139.9M, Input size=480 × 640, Flops=134.3G2026.0457.8——
ADBNetBackbone=ConvNext-base, Params.=299.1M, Input size=480 × 640, Flops=361.7G2026.0457.6——
SigmaBackbone=VMamba-S, Params.=69.8M, Input size=480 × 640, Flops=139.8G2026.0457——
CMXBackbone=MiT-B5, Params.=181.1M, Input size=480 × 640, Flops=167.8G2026.0456.9——
CMNextBackbone=MiT-B4, Params.=119.6M, Input size=480 × 640, Flops=131.9G2026.0456.9——
GeminiFusionBackbone=MiT-B3, Params.=75.8M, Input size=480 × 640, Flops=174.0G2026.0456.8——
RSGMamba-SBackbone=SegMAN-S, Params.=28.1M, Input size=480 × 640, Flops=58.5G2026.0456.4——
CMXBackbone=MiT-B4, Params.=139.9M, Input size=480 × 640, Flops=134.3G2026.0456.3——
DiffPixelFormerBackbone=MiT-B3, Params.=85.4M, Input size=480 × 640, Flops=154.8G2026.0456.3——
DFNetBackbone=MiT-B2, Params.=41.2M, Input size=480 × 640, Flops=35.1G2026.0456.1——
ADBNetBackbone=ConvNext-tiny, Params.=45.9M, Input size=480 × 640, Flops=52.7G2026.0456——
ECMRNBackbone=DFormer, Params.=68.6M, Input size=480 × 640, Flops=135.3G2026.0454.6——
TokenFusionBackbone=MiT-B3, Params.=45.9M, Input size=480 × 640, Flops=94.4G2026.0454.2——
SigmaBackbone=VMamba-T, Params.=48.3M, Input size=480 × 640, Flops=90.4G2026.0453.9——
DCANetBackbone=VMamba, Params.=123.8M, Input size=480 × 640, Flops=138.5G2026.0453.3——
RefineNet-LW-101Data=RGB2021.0743.6——
TD2-PSP50Data=RGB2021.0743.5——
CI-NetData=RGB2021.0742.6—72.7
3DGNNData=RGBD2021.0742——
D-CNNData=RGBD2021.0741——
ContextData=RGB2021.0740.6—70
Eigen et al.Data=RGB2021.0734.1—65.6
B-SegNetData=RGB2021.0732.4—68
Deng et al.Data=RGBD2021.0731.5—63.8
FCNData=RGB2021.0729.2—60
baselineBackbone=ResNet-502017.05—40.670.3
Context2017.05—40.670
Cross-stitch networkBackbone=VGG-162018.01—34.865
Cross-stitch networkBackbone=VGG-16-Shortcut2018.01—3565.1
loop1 w/ gt-depthBackbone=ResNet-50, Number of loops=1, Depth Usage=ground-truth depth2017.05—42.571.1
loop1 w/ pred-depthBackbone=ResNet-50, Number of loops=1, Depth Usage=predicted depth2017.05—42.771.2
loop1 w/o depthBackbone=ResNet-50, Number of loops=1, Depth Usage=none2017.05—41.970.6
loop2Backbone=ResNet-50, Number of loops=2, Depth Usage=predicted depth2017.05—43.171.3
loop2 (test-aug)Backbone=ResNet-50, Number of loops=2, Depth Usage=predicted depth, test-time augmentation=true2017.05—44.572.1
Multi-task baselineBackbone=VGG-16-Shortcut2018.01—33.664.4
Multiple-task baselineBackbone=VGG-162018.01—33.464.2
NDDR-CNNBackbone=VGG-162018.01—36.266.4
NDDR-CNN-ShortcutBackbone=VGG-16-Shortcut2018.01—36.767
RefineNet-Res101Backbone=ResNet-1012017.05—44.7—
RefineNet-Res152Backbone=ResNet-1522017.05—46.573.6
RefineNet-Res50Backbone=ResNet-502017.05—43.8—
Single-task baselineBackbone=VGG-162018.01—33.564.1
Single-task baselineBackbone=VGG-16-Shortcut2018.01—33.564.4
Sluice networkBackbone=VGG-162018.01—34.965.2
Sluice networkBackbone=VGG-16-Shortcut2018.01—35.365.3
w/ gt-depthBackbone=ResNet-50, Depth Usage=ground-truth depth2017.05—41.370.8
w/ pred-depthBackbone=ResNet-50, Depth Usage=predicted depth2017.05—41.871.1