Shape reconstruction on MedSDF (test)
2.314CDSkeletal Latent Diffusion
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Skeletal Latent DiffusionSub-sampling points=2,560 via FPS2026.03 | 2.314 | 0.447 | 1.136 | 98.24 | |
| GeM3DDecoder=implicit-field auto-encoder, Sub-sampling points=2,560 via FPS2026.03 | 2.672 | 0.643 | 1.26 | 97.86 | |
| DGCNNDecoder=FoldingNet, Sub-sampling points=2,560 via FPS2026.03 | 3.681 | 1.793 | 2.104 | 87.29 | |
| PointNet++Decoder=FoldingNet, Sub-sampling points=2,560 via FPS2026.03 | 3.811 | 1.849 | 2.045 | 85.97 | |
| Point-SSMDecoder=FoldingNet, Sub-sampling points=2,560 via FPS2026.03 | 3.876 | 2.012 | 2.336 | 85.3 | |
| PCTDecoder=FoldingNet, Sub-sampling points=2,560 via FPS2026.03 | 5.577 | 3.875 | 3.029 | 67.98 | |
| Diff-PCDDecoder=PointNet-based shape embedding, Sub-sampling points=2,560 via FPS2026.03 | 6.001 | 1.278 | 3.023 | 61.47 |