Part Segmentation on Pascal-Part-116 (test)
31.67mIoU (Unseen)PartCLIPSeg
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PartCLIPSegBackbone=ViT-B/16, Evaluation Protocol=Oracle-Obj2024.06 | 31.67 | 50.02 | 38.79 | |
| CAT-SegBackbone=ViT-B/16, Evaluation Protocol=Oracle-Obj2024.06 | 28.72 | 36.2 | 32.03 | |
| CLIPSegBackbone=ViT-B/16, Evaluation Protocol=Oracle-Obj2024.06 | 27.54 | 48.91 | 35.24 | |
| CAT-SegBackbone=ViT-B/16, Evaluation Protocol=Pred-All2024.06 | 25.42 | 28.17 | 26.72 | |
| PartCLIPSegBackbone=ViT-B/16, Evaluation Protocol=Pred-All2024.06 | 23.56 | 43.91 | 30.67 | |
| PartCATSegTraining Dataset=PartImageNet, Evaluation Mode=Oracle-Obj2025.01 | 22.88 | — | — | |
| PartCLIPSegTraining Dataset=PartImageNet, Evaluation Mode=Oracle-Obj2025.01 | 19.86 | — | — | |
| ZSSeg+Backbone=ResNet-50, Evaluation Protocol=Oracle-Obj2024.06 | 19.04 | 54.43 | 28.21 | |
| VLPartBackbone=ResNet-50, Evaluation Protocol=Oracle-Obj2024.06 | 18.7 | 42.61 | 25.99 | |
| PartCATSegTraining Dataset=PartImageNet, Evaluation Mode=Pred-All2025.01 | 18.21 | — | — | |
| CLIPSegTraining Dataset=PartImageNet, Evaluation Mode=Oracle-Obj2025.01 | 14.87 | — | — | |
| PartCLIPSegTraining Dataset=PartImageNet, Evaluation Mode=Pred-All2025.01 | 14.74 | — | — | |
| CLIPSegBackbone=ViT-B/16, Evaluation Protocol=Pred-All2024.06 | 13.27 | 27.79 | 17.96 | |
| CAT-SegTraining Dataset=PartImageNet, Evaluation Mode=Oracle-Obj2025.01 | 12.5 | — | — | |
| CLIPSegTraining Dataset=PartImageNet, Evaluation Mode=Pred-All2025.01 | 11.72 | — | — | |
| CAT-SegTraining Dataset=PartImageNet, Evaluation Mode=Pred-All2025.01 | 11.55 | — | — | |
| VLPartBackbone=ResNet-50, Evaluation Protocol=Pred-All2024.06 | 9.04 | 35.21 | 14.39 | |
| ZSSeg+Backbone=ResNet-50, Evaluation Protocol=Pred-All2024.06 | 3.38 | 38.05 | 6.2 |