Surface Normal Estimation on PASCAL ID Clean (test)
15.51RMSE (Degrees)FTP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FTPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 15.51 | -18.3 | 0.403 | |
| LLRDBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 15.54 | -18.11 | 0.279 | |
| L2-SPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 16.51 | -13.01 | 0.343 | |
| TPGMBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 18.17 | -4.24 | 0.616 | |
| AdapterBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 18.19 | -4.15 | 0.229 | |
| Vanilla FTBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 18.98 | 0 | 0.288 | |
| MARS-SPBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 19.01 | 0.18 | 0.313 | |
| BitFitBackbone=Swin-Tiny, Pre-trained=ImageNet21K, Decoder=Segformer2023.10 | 20.01 | 5.43 | 0.24 |