Instance Segmentation on CityScapes 64 x 128 (test)
52IoUFSNet+DeepLab
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FSNet+DeepLabDownsampling Strategy=FSNet, Backbone=DeepLabV3, Resolution=64 x 1282025.03 | 52 | 53 | |
| FSNet+SegFormer-B5Downsampling Strategy=FSNet, Backbone=SegFormer-B5, Resolution=64 x 1282025.03 | 51 | 52 | |
| FSNet+HRNetDownsampling Strategy=FSNet, Backbone=HRNet, Resolution=64 x 1282025.03 | 47 | 49 | |
| FSNet+SegFormer-B4Downsampling Strategy=FSNet, Backbone=SegFormer-B4, Resolution=64 x 1282025.03 | 46 | 48 | |
| LTDDownsampling Strategy=Learn-to-Downsample, Resolution=64 x 1282025.03 | 37 | 38 | |
| Avg+SegFormer-B5Downsampling Strategy=Uniform Subsampling, Backbone=SegFormer-B5, Resolution=64 x 1282025.03 | 27 | 29 | |
| Avg+DeepLabDownsampling Strategy=Uniform Subsampling, Backbone=DeepLabV3, Resolution=64 x 1282025.03 | 26 | 27 | |
| Avg+SegFormer-B4Downsampling Strategy=Uniform Subsampling, Backbone=SegFormer-B4, Resolution=64 x 1282025.03 | 25 | 27 | |
| Avg+HRNetDownsampling Strategy=Uniform Subsampling, Backbone=HRNet, Resolution=64 x 1282025.03 | 20 | 21 |