Single-view 3D reconstruction on Pix3D (test)
0.521IoUAutoSDF
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| AutoSDFvariant=Full2022.03 | 0.521 | 2.267 | 41.5 | — | — | — | |
| AutoSDFvariant=Sequential2022.03 | 0.516 | 2.254 | 41.2 | — | — | — | |
| ResNet2Voxel2022.03 | 0.505 | 4.67 | 35.7 | — | — | — | |
| Pix2Vox2022.03 | 0.504 | 3.001 | 38.5 | — | — | — | |
| Image-JE2022.03 | 0.48 | 2.983 | 39.4 | — | — | — | |
| ResNet2TSDF2022.03 | 0.475 | 4.582 | 35.1 | — | — | — | |
| LRGT†Transformer Blocks=Original self-attention blocks replace LGA transformer blocks2023.08 | 0.304 | — | — | 13 | — | — | |
| UMIFormerinput=single-view2023.02 | 0.3 | — | — | 12.9 | — | — | |
| UMIFormer2023.08 | 0.3 | — | — | 12.9 | — | — | |
| LRGT2023.08 | 0.299 | — | — | 12.7 | — | — | |
| 3D-RETRinput=single-view2023.02 | 0.297 | — | — | 12.5 | — | — | |
| 3D-RETR2023.08 | 0.297 | — | — | 12.5 | — | — | |
| GARNetinput=single-view2023.02 | 0.291 | — | — | 11.6 | — | — | |
| GARNet2023.08 | 0.291 | — | — | 11.6 | — | — | |
| Pix2Vox++input=single-view2023.02 | 0.279 | — | — | 11.3 | — | — | |
| Pix2Vox++2023.08 | 0.279 | — | — | 11.3 | — | — | |
| Cat3Drepresentation=implicit SDF, supervision=no 3D/viewpoint annotation2023.04 | — | 0.679 | — | 9.6 | 44.1 | 72.62 | |
| ShapeClipperconfiguration=Full model2023.04 | — | 0.618 | — | 13.17 | 54.73 | 80.02 | |
| ShapeClipper (w/o Lssc)semantic constraint=None, ablation=without CLIP-based semantic constraint2023.04 | — | 0.749 | — | 9.58 | 43.09 | 70.93 | |
| ShapeClipper (w/o noise-tol)outlier dropping=None, ablation=without outlier dropping in geometric constraint2023.04 | — | 0.64 | — | 12.77 | 53.19 | 78.61 | |
| ShapeClipper (w/o normal)geometric constraint=None, ablation=without normal-based geometric constraint2023.04 | — | 0.766 | — | 8.15 | 39.13 | 69.82 | |
| SSMPrepresentation=voxel prediction, supervision=single-view supervised voxel reconstruction2023.04 | — | 0.707 | — | 9.48 | 42.61 | 71.68 |