Instance Segmentation on Aria Everyday Activities 180 x 180 (test)
0.58mIoUFSNet+SegFormer-B5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FSNet+SegFormer-B5Downsampling Strategy=FSNet, Backbone=SegFormer-B5, Resolution=180 x 1802025.03 | 0.58 | 0.59 | |
| FSNet+HRNetDownsampling Strategy=FSNet, Backbone=HRNet, Resolution=180 x 1802025.03 | 0.57 | 0.58 | |
| FSNet+SegFormer-B4Downsampling Strategy=FSNet, Backbone=SegFormer-B4, Resolution=180 x 1802025.03 | 0.56 | 0.57 | |
| FSNet+DeepLabDownsampling Strategy=FSNet, Backbone=DeepLabV3, Resolution=180 x 1802025.03 | 0.54 | 0.56 | |
| LTDDownsampling Strategy=Learn-to-Downsample, Resolution=180 x 1802025.03 | 0.41 | 0.43 | |
| Avg+HRNetDownsampling Strategy=Uniform Subsampling, Backbone=HRNet, Resolution=180 x 1802025.03 | 0.39 | 0.41 | |
| Avg+SegFormer-B5Downsampling Strategy=Uniform Subsampling, Backbone=SegFormer-B5, Resolution=180 x 1802025.03 | 0.37 | 0.38 | |
| Avg+DeepLabDownsampling Strategy=Uniform Subsampling, Backbone=DeepLabV3, Resolution=180 x 1802025.03 | 0.36 | 0.37 | |
| Avg+SegFormer-B4Downsampling Strategy=Uniform Subsampling, Backbone=SegFormer-B4, Resolution=180 x 1802025.03 | 0.36 | 0.36 |