Human Parts Segmentation on PASCAL Human Parts ID Clean (test)
65.5mIoUFTP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FTPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.3972023.10 | 65.5 | 4.63 | |
| LLRDBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.2782023.10 | 64.37 | 2.81 | |
| TPGMBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.6022023.10 | 63.29 | 1.1 | |
| MARS-SPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.3082023.10 | 62.92 | 0.5 | |
| Vanilla FTBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.282023.10 | 62.61 | 0 | |
| L2-SPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.3362023.10 | 62.26 | -0.56 | |
| AdapterBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.2212023.10 | 60.84 | -2.81 | |
| BitFitBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Time (s/it)=0.2352023.10 | 59.06 | -5.67 |