Capsule endoscopy keyframe detection and diagnostic performance on VideoCAP (test)
44.12Lesion-Level Detection RateDiCE
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DiCESize=0.2B, Frames=29.98, Training Setting=Full training2026.04 | 44.12 | 85.29 | 19.3 | 54.86 | 77.81 | 20 | 67.5 | |
| Dinov3 + ViLAMPSize=0.2B, Frames=32, Training Setting=Full training, Category=General2026.04 | 35.29 | 61.76 | 23.18 | 69.69 | 83.26 | 12.5 | 60 | |
| Lingshu + ViLAMPSize=7B, Frames=32, Training Setting=Full training, Category=Medical2026.04 | 35.29 | 67.65 | 21.8 | 59.91 | 82.32 | 7.5 | 65 | |
| Colongpt + ViLAMPSize=2B, Frames=32, Training Setting=Full training, Category=Medical2026.04 | 34.56 | 63.97 | 21.86 | 65.21 | 85.07 | 12.5 | 57.5 | |
| Dinov3 + AKSSize=0.2B, Frames=32, Training Setting=Full training, Category=General2026.04 | 33.82 | 50 | 22.88 | 69.13 | 84.78 | 15 | 60 | |
| InternVL3.5 + ViLAMPSize=8B, Frames=32, Training Setting=Full training, Category=General2026.04 | 27.94 | 66.18 | 22.62 | 66.28 | 78.86 | 7.5 | 40 | |
| Lingshu + AKSSize=7B, Frames=32, Training Setting=Full training, Category=Medical2026.04 | 27.21 | 50.74 | 21.86 | 64.54 | 87.48 | 10 | 52.5 | |
| Colongpt + AKSSize=2B, Frames=32, Training Setting=Full training, Category=Medical2026.04 | 25 | 50.74 | 23.05 | 78.16 | 89.41 | 10 | 45 | |
| Qwen3-VL + ViLAMPSize=8B, Frames=32, Training Setting=Full training, Category=General2026.04 | 21.32 | 65.44 | 18.83 | 70.82 | 88.69 | 2.5 | 35 | |
| InternVL3.5 + AKSSize=8B, Frames=32, Training Setting=Full training, Category=General2026.04 | 19.85 | 51.47 | 23.55 | 67.77 | 83.45 | 5 | 32.5 | |
| Qwen3-VL + AKSSize=8B, Frames=32, Training Setting=Full training, Category=General2026.04 | 18.38 | 48.53 | 19.19 | 78.42 | 91.73 | 2.5 | 32.5 |