Part Segmentation on PartImageNet (val)
64.64mIoUCompositor
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Compositorbackbone=Swin-T, params=51M, training strategy=Models on parts and objects are trained jointly2023.06 | 64.64 | 78.31 | |
| MaskFormerbackbone=Swin-T, params=92M (46M×2), training strategy=Models on parts and objects are trained separately2023.06 | 63.96 | 77.37 | |
| UniLSegSA-1B training data scale=100%2023.12 | 63.87 | — | |
| UniLSegSA-1B training data scale=20%2023.12 | 62.46 | — | |
| SegFormerbackbone=MiT-B2, params=48M (24M×2), training strategy=Models on parts and objects are trained separately2023.06 | 61.97 | 73.77 | |
| SegFormer2023.12 | 61.97 | — | |
| MaskFormer-Dualbackbone=Swin-T, params=51M, training strategy=Models on parts and objects are trained jointly2023.06 | 61.69 | 75.64 | |
| Compositorbackbone=ResNet-50, params=50M, training strategy=Models on parts and objects are trained jointly2023.06 | 61.44 | 73.41 | |
| Deeplab v3+backbone=ResNet-50, params=84M (42M×2), training strategy=Models on parts and objects are trained separately2023.06 | 60.57 | 71.07 | |
| DeepLabV3+2023.12 | 60.57 | — | |
| MaskFormerbackbone=ResNet-50, params=90M (45M×2), training strategy=Models on parts and objects are trained separately2023.06 | 60.34 | 72.75 | |
| MaskFormer-Dualbackbone=ResNet-50, params=50M, training strategy=Models on parts and objects are trained jointly2023.06 | 58.02 | 70.42 | |
| SemanticFPN2023.12 | 56.76 | — |