List your company
ResearchFormerly WizWand

Semantic Segmentation on NYUD v2 (evaluation)

55.8mIoUOurs (w SMMCL)
50.91252.18153.4554.719Aug 23, 2023
Updated 8mo ago

Evaluation results

MethodLinks
Ours (w SMMCL)Backbone=SegNeXt-B, Multi-scale evaluation=false, SMMCL=true2023.0855.8
Ours (w SMMCL)Backbone=SegFormer-B2, Multi-scale evaluation=false, SMMCL=true2023.0855.1
TokenFusionBackbone=SegFormer-B3, Multi-scale evaluation=false2023.0854.8
Base (w/o SMMCL)Backbone=SegNeXt-B, Multi-scale evaluation=false, SMMCL=false2023.0854.7
CMXBackbone=SegFormer-B2, Multi-scale evaluation=false2023.0854.1
Base (w/o SMMCL)Backbone=SegFormer-B2, Multi-scale evaluation=false, SMMCL=false2023.0853.7
Ours (w SMMCL)Backbone=ResNet-101, Multi-scale evaluation=false, SMMCL=true2023.0853.4
TokenFusionBackbone=SegFormer-B2, Multi-scale evaluation=false2023.0853.3
Base (w/o SMMCL)Backbone=ResNet-101, Multi-scale evaluation=false, SMMCL=false2023.0852.5
SA-GateBackbone=ResNet-101, Multi-scale evaluation=true2023.0852.4
ShapeConvBackbone=ResNeXt-101, Multi-scale evaluation=true2023.0851.3
CENBackbone=ResNet-101, Multi-scale evaluation=false2023.0851.1