Medical Image Segmentation on Sites A-E Average
81.6Dice ScoreMCDRL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MCDRLBackbone=ViT-L/14, Pre-trained weights=CLIP2025.08 | 81.6 | 74.3 | 94.3 | |
| MCDRLBackbone=ViT-B/16, Pre-trained weights=CLIP2025.08 | 80 | 72.9 | 93.5 | |
| BiomedCoOpBackbone=ViT-L/14, Pre-trained weights=CLIP2025.08 | 79.4 | 72.4 | 93.2 | |
| MCDRLBackbone=ResNet-50, Pre-trained weights=CLIP2025.08 | 78.6 | 71.5 | 92.6 | |
| StyLIPBackbone=ViT-L/14, Pre-trained weights=CLIP2025.08 | 78.3 | 71.2 | 92.5 | |
| BiomedCoOpBackbone=ViT-B/16, Pre-trained weights=CLIP2025.08 | 78 | 71.1 | 92.3 | |
| StyLIPBackbone=ViT-B/16, Pre-trained weights=CLIP2025.08 | 76.8 | 69.9 | 91.6 | |
| BiomedCoOpBackbone=ResNet-50, Pre-trained weights=CLIP2025.08 | 76.7 | 70.1 | 91.4 | |
| StyLIPBackbone=ResNet-50, Pre-trained weights=CLIP2025.08 | 75.4 | 68.6 | 90.7 | |
| BaselineBackbone=ViT-L/14, Pre-trained weights=CLIP2025.08 | 75.1 | 68.3 | 90.9 | |
| BaselineBackbone=ViT-B/16, Pre-trained weights=CLIP2025.08 | 73.6 | 66.9 | 90 | |
| BaselineBackbone=ResNet-50, Pre-trained weights=CLIP2025.08 | 72.1 | 65.6 | 88.9 |