Kubric
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Visual World Modelling | Kubric | GPT-4o Score7.45 | 18 | |
| Point Tracking | Kubric | AJ74 | 18 | |
| Future Surface Normals Estimation | Kubric (test) | a393.45 | 12 | |
| Future Depth Estimation | Kubric (test) | d1 Score88.8 | 12 | |
| Future Semantic Segmentation | Kubric (test) | mIoU (All Classes)70.55 | 12 | |
| Dense Tracking | Kubric | EPE1.525 | 11 | |
| Forward-dynamics Prediction | Kubric AURORA-BENCH | GPT-4o Score7.32 | 9 | |
| Point Tracking | Kubric TAP-Vid | Average Jaccard88.3 | 9 | |
| 4D reconstruction | Kubric G | EPE P0 (t1)0.024 | 8 | |
| 4D reconstruction | Kubric-F | EPE P0 (t1)0.019 | 8 | |
| Novel View Synthesis | Kubric-4D-dyn dynamic only (target 3) | PSNR17.37 | 7 | |
| Novel View Synthesis | Kubric-4D-dyn dynamic only (target 2) | PSNR19.1 | 7 | |
| Novel View Synthesis | Kubric-4D-dyn dynamic only (target 1) | PSNR22.27 | 7 | |
| World-centric geometry reconstruction | Kubric | Rel^p3.4 | 7 | |
| Point Tracking | Kubric (test) | Mean EPE2.51 | 7 | |
| Point Tracking | Kubric TAP-Vid (test) | AJ40.1 | 7 | |
| Dense Scene Flow Estimation | Kubric-4D Static Camera Scene Flow (held-out scenes) | EPE0.1 | 6 | |
| Novel View Synthesis | Kubric Pillow | LPIPS0.018 | 6 | |
| Novel View Synthesis | Kubric Bag | LPIPS0.006 | 6 | |
| Novel View Synthesis | Kubric Cars | LPIPS0.035 | 6 | |
| Novel View Synthesis | Kubric Car | LPIPS0.013 | 6 | |
| Video-to-4D Scene Generation | Kubric 14 (test) | CLIP Similarity86.41 | 6 | |
| Trajectory Reconstruction | Kubric | VEPE0.75 | 4 | |
| Action-centric Editing | Kubric (test) | Human Evaluation Score0.4 | 4 | |
| Dense Tracking | Kubric G | EPE0.035 | 4 |
Showing 25 of 28 rows