Pytorch unofficial implementation of Dis-PU
(Point Cloud Upsampling via Disentangled Refinement)
https://arxiv.org/abs/2106.04779
Model
Detail
Input : 2048
Output : 8192
Test Dataset : PU-GAN dataset
| X4 | Chamfer Distance(10-3) | HD(10-3) | P2F(10-3) |
|---|---|---|---|
| This code | 0.2702 | 5.50 | 4.35 |
| Paper | 0.315 | 4.201 | 4.141 |
Ground Truth
Model Prediction
This repo is heavily built on Point Transformer code.