ScanNet++
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Semantic Segmentation | ScanNet++ (val) | mIoU52.9 | 32 | |
| Multi-view 3D Reconstruction | ScanNet++ (test) | AUC@5cm59 | 30 | |
| Indoor Semantic Segmentation | ScanNet++ (val) | mIoU50.7 | 20 | |
| 3D Semantic Segmentation | ScanNet++ (val) | mAcc55.8 | 16 | |
| 3D Line Mapping | ScanNet++ 30 scenes 16 | Accuracy Line (ACC-L)0.1337 | 16 | |
| 3D line mapping | ScanNet++ 16 (30 scenes) | Recall@514.31 | 16 | |
| TSDF Reconstruction | ScanNet++ | Accuracy15.67 | 15 | |
| Semantic Segmentation | ScanNet++ | Mean IoU (mIoU)62.4 | 15 | |
| Panoptic Segmentation | ScanNet++ | PQ (Panoptic Quality)77.09 | 14 | |
| Multi-view Camera Pose Estimation | ScanNet++ | RPA @ 5°2,270 | 14 | |
| 3D instance segmentation | ScanNet++ V1 (val) | AP5035.8 | 12 | |
| Video depth estimation | ScanNet++ | Absolute Relative Error4.4 | 10 | |
| 3D Object Detection | ScanNet++ | mAP@0.2581.06 | 10 | |
| 3D Reconstruction | ScanNet++ | F1 Score41.1 | 8 | |
| 3D Gaussian Splatting Compression | ScanNet++ 0.2 compression ratio (test) | LPIPS0.1948 | 8 | |
| 3D Semantic Segmentation | ScanNet ➜ ScanNet++ 20 classes v2 (test) | mIoU51.66 | 8 | |
| 3D Semantic Segmentation | ScanNet++ ➜ ScanNet (19 classes) v2 (test) | mIoU50.8 | 8 | |
| Instance Segmentation | ScanNet++ 52 (val) | mAP@0.2546.3 | 8 | |
| 3D Semantic Segmentation | ScanNet++ 100 classes (test) | f-mIoU29.6 | 8 | |
| Monocular Depth Estimation | ScanNet++ (val) | Relative Error (Rel)0.2169 | 8 | |
| Class-Agnostic 3D Instance Segmentation | ScanNet++ 1554 classes 1.0 (val) | AP22.2 | 8 | |
| 3D Instance Segmentation | ScanNet++ | AP@2521.7 | 7 | |
| Novel view synthesis | ScanNet++ 512v (test) | PSNR20.308 | 7 | |
| Novel view synthesis | ScanNet++ 256v (test) | PSNR20.797 | 7 | |
| Class-agnostic 3D Instance Segmentation | ScanNet++ (test) | AP22.2 | 7 |
Showing 25 of 65 rows