Semantic Segmentation on BDD100K (test)
66.1mIoUDeDelayed
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeDelayedLocal delay=0 ms, Remote delay=0 ms2025.10 | 66.1 | — | — | |
| DeDelayedLocal delay=0 ms, Remote delay=33 ms2025.10 | 66 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=0 ms2025.10 | 65.6 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=0 ms2025.10 | 65.5 | — | — | |
| DeDelayedLocal delay=0 ms, Remote delay=67 ms2025.10 | 65.5 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=33 ms2025.10 | 65.5 | — | — | |
| DeDelayedLocal delay=0 ms, Remote delay=100 ms2025.10 | 65.2 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=0 ms2025.10 | 65.2 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=67 ms2025.10 | 65 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=33 ms2025.10 | 65 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=33 ms2025.10 | 64.9 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=100 ms2025.10 | 64.7 | — | — | |
| DSNet-BaseModel Category=Non-real-time model, #Params=37.5M, Resolution=1280 x 7202024.06 | 64.6 | — | — | |
| DeDelayedLocal delay=0 ms, Remote delay=133 ms2025.10 | 64.6 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=67 ms2025.10 | 64.6 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=100 ms2025.10 | 64.3 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=133 ms2025.10 | 64.2 | — | — | |
| DeDelayedLocal delay=0 ms, Remote delay=167 ms2025.10 | 64.1 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=67 ms2025.10 | 64 | — | — | |
| DeDelayedLocal delay=4 ms, Remote delay=167 ms2025.10 | 63.7 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=133 ms2025.10 | 63.7 | — | — | |
| Deeplabv3+Model Category=Non-real-time model, #Params=40.5M, Resolution=1280 x 7202024.06 | 63.6 | — | — | |
| DeDelayedLocal delay=8 ms, Remote delay=167 ms2025.10 | 63.2 | — | — | |
| DANetModel Category=Non-real-time model, #Params=48.1M, Resolution=1280 x 7202024.06 | 62.8 | — | — | |
| DSNet-head64Model Category=Real-time model, #FPS=172.2, #Params=6.6M, GPU=RTX4090, Resolution=1280 x 7202024.06 | 62.8 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=100 ms2025.10 | 62.7 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=0 ms2025.10 | 62.4 | — | — | |
| PSPNetModel Category=Non-real-time model, #Params=31.1M, Resolution=1280 x 7202024.06 | 62.3 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=33 ms2025.10 | 61.9 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=67 ms2025.10 | 61.6 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=100 ms2025.10 | 61.2 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=133 ms2025.10 | 61.1 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=133 ms2025.10 | 60.7 | — | — | |
| SFNet(Res18)Model Category=Real-time model, #FPS=132.5, #Params=12.9M, GPU=RTX4090, Resolution=1280 x 7202024.06 | 60.6 | — | — | |
| SFNet-Lite(Res18)Model Category=Real-time model, #FPS=161.3, #Params=12.3M, GPU=RTX4090, Resolution=1280 x 7202024.06 | 60.6 | — | — | |
| DeDelayedLocal delay=33 ms, Remote delay=167 ms2025.10 | 60.3 | — | — | |
| SFNet(DF2)Model Category=Real-time model, #FPS=208.2, #Params=19.6M, GPU=RTX4090, Resolution=1280 x 7202024.06 | 60.2 | — | — | |
| OCRNetModel Category=Non-real-time model, #Params=39.0M, Resolution=1280 x 7202024.06 | 60.1 | — | — | |
| Remote (video-predictive)Local delay=–, Remote delay=167 ms2025.10 | 59.6 | — | — | |
| SFNet-Lite(STDC2)Model Category=Real-time model, #FPS=194.5, #Params=13.7M, GPU=RTX4090, Resolution=1280 x 7202024.06 | 59.4 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=0 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=33 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=67 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=100 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=133 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=0 ms, Remote delay=167 ms2025.10 | 58.8 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=0 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=33 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=67 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=100 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=133 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=4 ms, Remote delay=167 ms2025.10 | 58.5 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=0 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=33 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=67 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=100 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=133 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=8 ms, Remote delay=167 ms2025.10 | 58.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=0 ms2025.10 | 56.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=33 ms2025.10 | 56.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=67 ms2025.10 | 56.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=100 ms2025.10 | 56.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=133 ms2025.10 | 56.2 | — | — | |
| Local (image only)Local delay=33 ms, Remote delay=167 ms2025.10 | 56.2 | — | — | |
| CMFormerBackbone=Swin-B, Trained on=G+S2023.07 | 53.36 | — | — | |
| Pin the MemoryBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 50.18 | — | — | |
| RobustNetBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 49 | — | — | |
| IBN-NetBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 48.91 | — | — | |
| MLDGBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 48.52 | — | — | |
| BaselineBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 47.47 | — | — | |
| TSMLDGBackbone=ResNet-50, Architecture=DeepLabV3+, Source Datasets=GTAV + Synthia + IDD2022.04 | 46.43 | — | — | |
| SHADEBackbone=ResNet-101, Source Dataset=GTAV, Extra Data=false, Checkpoint Selection=Last2022.04 | 43.66 | — | — | |
| SPCBackbone=ResNet-50, Trained on=G+S2023.07 | 43.18 | — | — | |
| FSDRBackbone=ResNet-101, Source Dataset=GTAV, Extra Data=true, Checkpoint Selection=Best2022.04 | 41.2 | — | — | |
| ReVT {1,4,6}Encoder=MiT-B5, Parameters=84.7M, Single Source=true, Frame Rate (fps)=9.72023.08 | 40.3 | — | — | |
| SHADEBackbone=ResNet-50, Trained on=G+S2023.07 | 40.3 | — | — | |
| ReVT {4,5,6}Encoder=MiT-B5, Parameters=84.7M, Single Source=true, Frame Rate (fps)=9.72023.08 | 39.62 | — | — | |
| SHADEBackbone=ResNet-50, Source Dataset=GTAV, Extra Data=false, Checkpoint Selection=Last2022.04 | 39.28 | — | — | |
| AdvStyleBackbone=ResNet-50, Trained on=G+S2023.07 | 39.26 | — | — | |
| ReVT {4,5,6}Encoder=MiT-B3, Parameters=47.2M, Single Source=true, Frame Rate (fps)=10.72023.08 | 38.73 | — | — | |
| DRPCBackbone=ResNet-101, Source Dataset=GTAV, Extra Data=true, Checkpoint Selection=Best2022.04 | 38.72 | — | — | |
| ReVT {1,4,6}Encoder=MiT-B3, Parameters=47.2M, Single Source=true, Frame Rate (fps)=10.72023.08 | 38.65 | — | — | |
| GMDGObjective=GAim1+GReg1+GAim2+GReg22024.02 | 38.62 | — | — | |
| GAim1+GAim2+GReg2Objective=GAim1+GAim2+GReg22024.02 | 38.58 | — | — | |
| GReg1+GAim2+GReg2Objective=GReg1+GAim2+GReg22024.02 | 38.13 | — | — | |
| RobustNetObjective=GAim2, GReg22024.02 | 37.69 | — | — | |
| FSDREncoder=ResNet-101, Parameters=68.6M, Single Source=false, Frame Rate (fps)=5.32023.08 | 37.4 | — | — | |
| SAN+SAWEncoder=ResNet-101, Parameters=44.6M, Single Source=true, Frame Rate (fps)=5.32023.08 | 35.98 | — | — | |
| IBN-Net2024.02 | 35.55 | — | — | |
| DeepLabv3+2024.02 | 35.46 | — | — | |
| SAN+SAWEncoder=ResNet-50, Parameters=25.6M, Single Source=true, Frame Rate (fps)=8.12023.08 | 35.42 | — | — | |
| ISWBackbone=ResNet-50, Source Dataset=GTAV, Extra Data=false, Checkpoint Selection=Last2022.04 | 35.2 | — | — | |
| BaselineEncoder=MiT-B5, Parameters=84.7M, Single Source=true, Frame Rate (fps)=9.72023.08 | 35.19 | — | — | |
| ReVT {4,5,6}Encoder=MiT-B2, Parameters=27.4M, Single Source=true, Frame Rate (fps)=122023.08 | 35.18 | — | — | |
| ReVT {1,4,6}Encoder=MiT-B2, Parameters=27.4M, Single Source=true, Frame Rate (fps)=122023.08 | 34.53 | — | — | |
| DRPCEncoder=ResNet-101, Parameters=68.6M, Single Source=false, Frame Rate (fps)=5.32023.08 | 34.34 | — | — | |
| IBN-NetBackbone=ResNet-101, Source Dataset=GTAV, Extra Data=false, Checkpoint Selection=Last2022.04 | 34.21 | — | — | |
| ISWBackbone=ResNet-50, Trained on=G+S2023.07 | 34.09 | — | — | |
| ISWBackbone=ResNet-101, Source Dataset=GTAV, Extra Data=false, Checkpoint Selection=Last2022.04 | 33.36 | — | — | |
| BaselineEncoder=MiT-B3, Parameters=47.2M, Single Source=true, Frame Rate (fps)=10.72023.08 | 33.33 | — | — |