Surgical Triplet Recognition on CholecT50
80.53APIMIS-DINOv2
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MIS-DINOv2Backbone=ViT-L, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 80.53 | 53.48 | 40.9 | 30.72 | 33.48 | 26.34 | — | |
| ZENBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 80.48 | 52.53 | 41.7 | 30.12 | 34.01 | 26.54 | — | |
| MIS-DINOv2Backbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 69.8 | 46.98 | 35.54 | 25.71 | 27.77 | 21.46 | — | |
| SurgeNetBackbone=CAFormer18, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 61.1 | 43.63 | 34.72 | 22.84 | 24.2 | 18.97 | — | |
| MIS-DINOBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 59.99 | 41.22 | 35.9 | 22.58 | 24.89 | 19.86 | — | |
| PeskaVLPBackbone=ResNet50, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 56.24 | 39.92 | 26.46 | 20.45 | 16.76 | 13.21 | — | |
| ImageNetBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 52.22 | 35.2 | 25.09 | 18.51 | 15.36 | 11.12 | — | |
| MIS-MoCoV3Backbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 46.29 | 29.93 | 23.57 | 14.82 | 12.19 | 8.76 | — | |
| EndoFMBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 43.15 | 30.95 | 23.62 | 15.49 | 14.19 | 10.99 | — | |
| MIS-MSNBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 38.69 | 27.94 | 25.46 | 14.17 | 13.41 | 10.44 | — | |
| EndoViTBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 38.17 | 27.12 | 20.33 | 13.81 | 11.97 | 9.15 | — | |
| MIS-MAEBackbone=ViT-B, Frozen Backbones=true, Pre-training protocol=Self-supervised learning2026.02 | 37.7 | 26.45 | 21.78 | 13.54 | 12.39 | 9.27 | — | |
| GSViTBackbone=EfficientViT, Frozen Backbones=true, Pre-training protocol=Existing model2026.02 | 26.24 | 16.68 | 12.09 | 8.38 | 5.81 | 4.59 | — | |
| CliPPER (Ours-All)Zero-shot=true, Prompt averaging=4, Training Data=SVL + YouTube2026.03 | — | — | — | — | — | — | 6.6 | |
| CliPPER (Ours-SVL)Zero-shot=true, Prompt averaging=4, Training Data=SVL2026.03 | — | — | — | — | — | — | 5.2 | |
| CliPPER (Ours-YT)Zero-shot=true, Prompt averaging=4, Training Data=YouTube2026.03 | — | — | — | — | — | — | 5.8 | |
| HecVLZero-shot=true, Prompt averaging=42026.03 | — | — | — | — | — | — | 3.5 | |
| PeskaVLPZero-shot=true, Prompt averaging=42026.03 | — | — | — | — | — | — | 4.3 | |
| SurgVLPZero-shot=true, Prompt averaging=42026.03 | — | — | — | — | — | — | 3.7 | |
| VindLUZero-shot=true, Prompt averaging=42026.03 | — | — | — | — | — | — | 3 | |
| VindLU-SVLZero-shot=true, Prompt averaging=42026.03 | — | — | — | — | — | — | 5 |