Unconditional 3D Point Cloud Generation on ShapeNet Car v1 (test)
0.901MMD (CD)PointFlow
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PointFlow2022.12 | 0.901 | 0.8071 | 46.88 | 50 | |
| PSFSampling steps (N)=12022.12 | 1.023 | 0.802 | 42.89 | 53.12 | |
| PVD2022.12 | 1.077 | 0.7938 | 41.19 | 50.56 | |
| DPF-Net2022.12 | 1.129 | 0.8529 | 45.74 | 49.43 | |
| SoftFlow2022.12 | 1.187 | 0.8594 | 42.9 | 44.6 | |
| PVD-DDIMSampling protocol=DDIM, Sampling steps (N)=1002022.12 | 1.202 | 0.8176 | 40.01 | 48.34 | |
| l-GAN (CD)Training metric=CD2022.12 | 1.532 | 1.226 | 38.92 | 23.58 | |
| Shape-GF2022.12 | 9.232 | 0.7558 | 49.43 | 50.28 |