Semantic Segmentation on MaSS13K 500 (val)
88.97mIoUMaSSFormer
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MaSSFormerBackbone=R50, Param=37.42M, FLOPs=2036G, Input Resolution=4096 x 40962025.03 | 88.97 | 48.97 | 56.39 | |
| Mask2FormerBackbone=R50, Param=44.01M, FLOPs=3123G, Input Resolution=4096 x 40962025.03 | 88.28 | 47.4 | 54.58 | |
| MPFormerBackbone=R50, Param=43.9M, FLOPs=4155G, Input Resolution=4096 x 40962025.03 | 87.76 | 47.81 | 55.13 | |
| SegNeXtBackbone=MSCAN-B, Param=27.57M, FLOPs=1536G, Input Resolution=4096 x 40962025.03 | 87.71 | 39.93 | 46.15 | |
| FeedFormerBackbone=lvt, Param=4.65M, FLOPs=300G, Input Resolution=4096 x 40962025.03 | 87.17 | 42.06 | 48.38 | |
| MaSSFormer-LiteBackbone=R18, Param=15.07M, FLOPs=771G, Input Resolution=4096 x 40962025.03 | 87.11 | 45.35 | 51.37 | |
| OCRNetBackbone=R50, Param=36.52M, FLOPs=7352G, Input Resolution=4096 x 40962025.03 | 86.99 | 32.68 | 38.08 | |
| SeaFormer-LParam=13.95M, FLOPs=303G, Input Resolution=4096 x 40962025.03 | 86.78 | 38.61 | 44.98 | |
| DeepLabv3+Backbone=R50, Param=41.22M, FLOPs=8008G, Input Resolution=4096 x 40962025.03 | 86.66 | 40.08 | 47.18 | |
| PEMBackbone=R50, Param=35.5M, FLOPs=1859G, Input Resolution=4096 x 40962025.03 | 83.41 | 40.51 | 46.75 | |
| MaskFormerBackbone=R50, Param=41.31M, FLOPs=2396G, Input Resolution=4096 x 40962025.03 | 83.27 | 38.61 | 43.99 | |
| STDC2Param=12.30M, FLOPs=564G, Input Resolution=4096 x 40962025.03 | 83.08 | 27.99 | 33.34 | |
| PIDNet-LParam=37.08M, FLOPs=1653G, Input Resolution=4096 x 40962025.03 | 82.28 | 31.3 | 34.75 | |
| UperNetBackbone=R50, Param=64.04M, FLOPs=11373G, Input Resolution=4096 x 40962025.03 | 82.03 | 35.85 | 41.81 | |
| CGRSeg-LBackbone=EFv2-L, Param=35.64M, FLOPs=1536G, Input Resolution=4096 x 40962025.03 | 81.29 | 34.48 | 40.9 | |
| BiSeNetv2Param=3.35M, FLOPs=591G, Input Resolution=4096 x 40962025.03 | 71.55 | 25.05 | 31.82 |