Semantic Segmentation on NightCity+ (val)
62.82mIoUNightLab-HDM
Evaluation results
| Method | Links | |
|---|---|---|
| NightLab-HDMBackbone=Swin-Base, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 62.82 | |
| Night Lab-HDMAdaptation Approach=Dual-level segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Joint train with Cityscapes2022.04 | 62.82 | |
| NightLab-RDNBackbone=Swin-Base, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 62.11 | |
| Night Lab-RDNAdaptation Approach=Dual-level segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Joint train with Cityscapes2022.04 | 62.11 | |
| NightLab-HDMBackbone=Swin-Base, Resolution=1024x2048, Training Data=Night images only2022.04 | 60.73 | |
| Night Lab-HDMAdaptation Approach=Dual-level segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Single train2022.04 | 60.73 | |
| NightLab (DeeplabV3+)Backbone=Res101, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 60.41 | |
| NightLab-BaselineBackbone=Swin-Base, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 60.37 | |
| Night Lab-BAdaptation Approach=Segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Joint train with Cityscapes2022.04 | 60.37 | |
| NightLab-RDNBackbone=Swin-Base, Resolution=1024x2048, Training Data=Night images only2022.04 | 60.27 | |
| Night Lab-RDNAdaptation Approach=Dual-level segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Single train2022.04 | 60.27 | |
| UPer-SwinBackbone=Swin-Base, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 59.67 | |
| UPerNetBackbone=Swin-Base, Resolution=512x1024, Training Data=Cityscapes + Night images2022.04 | 59.35 | |
| UPerNetAdaptation Approach=Segmentation, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 59.35 | |
| NightLab-BaselineBackbone=Swin-Base, Resolution=1024x2048, Training Data=Night images only2022.04 | 59.25 | |
| Night Lab-BAdaptation Approach=Segmentation, Network=UPerNet-Swin-DeformConv, Training Protocol=Single train2022.04 | 59.25 | |
| DeeplabV3+Backbone=Res101, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 59.03 | |
| SingleHDRAdaptation Approach=Image Enhancement, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 58.88 | |
| DANNetAdaptation Approach=Network Adaptation, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 58.69 | |
| HRNetV2Backbone=HRNet-W48, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 58.49 | |
| DeeplabV3+Backbone=Res101, Resolution=512x1024, Training Data=Cityscapes + Night images2022.04 | 58.29 | |
| AdaptSegAdaptation Approach=Network Adaptation, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 58.29 | |
| UPer-SwinBackbone=Swin-Base, Resolution=1024x2048, Training Data=Night images only2022.04 | 58.25 | |
| UPer-ViTBackbone=ViT, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 58.07 | |
| DANetBackbone=Res101, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 57.72 | |
| UPerNetBackbone=Swin-Base, Resolution=512x1024, Training Data=Night images only2022.04 | 57.71 | |
| UPerNetAdaptation Approach=Segmentation, Network=UPerNet-Swin, Training Protocol=Single train2022.04 | 57.71 | |
| PSPNetBackbone=Res101, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 57.52 | |
| UPer-ViTBackbone=ViT, Resolution=1024x2048, Training Data=Night images only2022.04 | 57.13 | |
| SingleHDRAdaptation Approach=Image Enhancement, Network=UPerNet-Swin, Training Protocol=Single train2022.04 | 57.07 | |
| UPer-SwinBackbone=Res101, Resolution=1024x2048, Training Data=Cityscapes + Night images2022.04 | 56.98 | |
| PSPNetBackbone=Res101, Resolution=512x1024, Training Data=Cityscapes + Night images2022.04 | 56.89 | |
| NightLab (DeeplabV3+)Backbone=Res101, Resolution=1024x2048, Training Data=Night images only2022.04 | 56.21 | |
| DANetBackbone=Res101, Resolution=1024x2048, Training Data=Night images only2022.04 | 55.98 | |
| HRNetV2Backbone=HRNet-W48, Resolution=1024x2048, Training Data=Night images only2022.04 | 55.89 | |
| UPer-SwinBackbone=Res101, Resolution=1024x2048, Training Data=Night images only2022.04 | 55.81 | |
| PSPNetBackbone=Res101, Resolution=1024x2048, Training Data=Night images only2022.04 | 55.64 | |
| PSPNetBackbone=Res101, Resolution=512x1024, Training Data=Night images only2022.04 | 54.75 | |
| DeeplabV3+Backbone=Res101, Resolution=1024x2048, Training Data=Night images only2022.04 | 54.47 | |
| DeeplabV3+Backbone=Res101, Resolution=512x1024, Training Data=Night images only2022.04 | 54.21 | |
| NightCityAdaptation Approach=Exposure-Aware, Network=Res101, Training Protocol=Joint train with Cityscapes2022.04 | 53.9 | |
| NightCityAdaptation Approach=Exposure-Aware, Network=Res101, Training Protocol=Single train2022.04 | 51.8 | |
| CycleGANAdaptation Approach=Image Translation, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 44.07 | |
| Pix2PixHDAdaptation Approach=Image Translation, Network=UPerNet-Swin, Training Protocol=Joint train with Cityscapes2022.04 | 43.38 |