This is implementation of PCN——Point Completion Network in pytorch. PCN is an autoencoder for point cloud completion. As for the details of the paper, please refer to arXiv.
- Python 3.7.9
- PyTorch 1.7.0
- CUDA 10.1.243
I just use the category of car to train, evaluation and test the model. All of the data is created by myself. By running the evaluation script evaluation.py
, we can get a average cd loss 0.0097
on the test split.