Mask Classification on MESS Benchmark
68.6General ScoreMaskCLIP++
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MaskCLIP++Architecture=ViT-L/14, Input Resolution=392 (short side), Training Protocol=Finetuned on COCO2024.12 | 68.6 | 74.7 | 50.3 | — | 65.5 | |
| CAT-SegArchitecture=ViT-L/14, Input Resolution=392 (short side), Training Protocol=Finetuned on COCO2024.12 | 60.7 | 66.9 | 62 | 34.9 | 64.4 | |
| EVA02 CLIPArchitecture=ViT-L/14, Input Resolution=392 (short side), Training Protocol=Pretrained2024.12 | 56.9 | 60.5 | 61.7 | 33.8 | 52.4 | |
| OpenAI CLIPArchitecture=ViT-L/14, Input Resolution=392 (short side), Training Protocol=Pretrained2024.12 | 54.2 | 61.1 | 61 | 36.7 | 51 | |
| CLIPSelfArchitecture=ViT-L/14, Input Resolution=392 (short side), Training Protocol=Finetuned on COCO2024.12 | 52.5 | 45.7 | 55.5 | 45.3 | 56.3 |