Auto-Vocabulary 3D Object Detection on ScanNet V2 (val)
9.23mAPAV3DOD
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AV3DODSetting=joint-training2025.12 | 9.23 | 20.04 | 7.07 | 27.2 | 47.98 | 23.04 | 0.89 | 0.797 | 0.812 | 0.585 | 0.709 | 0.475 | 0.57 | |
| CoDASetting=joint-training2025.12 | 5.75 | 16.87 | 3.52 | 18.82 | 40.21 | 14.55 | 0.885 | 0.68 | 0.804 | 0.448 | 0.602 | 0.36 | 0.458 | |
| 3DETR+Florence2Setting=joint-training2025.12 | 5.16 | 18.05 | 2.58 | 15.87 | 40.53 | 10.94 | 0.876 | 0.651 | 0.806 | 0.421 | 0.57 | 0.339 | 0.433 | |
| 3DETR+Qwen3Setting=joint-training2025.12 | 4.95 | 16.31 | 2.67 | 15.49 | 36.57 | 11.28 | 0.911 | 0.622 | 0.801 | 0.395 | 0.567 | 0.316 | 0.418 | |
| 3DETR+PointLLMSetting=joint-training2025.12 | 3.8 | 13.84 | 1.79 | 12.75 | 25.89 | 10.12 | 0.875 | 0.612 | 0.775 | 0.384 | 0.536 | 0.298 | 0.394 | |
| 3DETR+Qwen3Setting=zero-shot2025.12 | 3.18 | 11.85 | 3.18 | 9.93 | 22.59 | 7.39 | 0.887 | 0.618 | 0.749 | 0.369 | 0.548 | 0.276 | 0.387 | |
| 3DETR+PointLLMSetting=zero-shot2025.12 | 2.84 | 14.89 | 0.43 | 6.88 | 32.77 | 1.71 | 0.896 | 0.616 | 0.754 | 0.369 | 0.552 | 0.278 | 0.389 | |
| 3DETR+Florence2Setting=zero-shot2025.12 | 2.83 | 13.82 | 0.63 | 8.06 | 28 | 4.07 | 0.867 | 0.616 | 0.749 | 0.371 | 0.534 | 0.278 | 0.382 |