PointCloudVAE

valiational autoencoder of pointcloud
CAUTION:
Changes may be made to test various things.

Environmet

• python 3.8.10
• cuda 11.6
• pytorch 1.32.1
• pytorch3d 0.7.2
• open3d 0.13.0

Dataset

ShapeNet Dataset from [[url]]

Usage

training

python train.py

evaluation

python eval.py -d "directory name of tar path which you want to test"

Settings for training and evaluation are in options.py.

Result

AutoEncoder

ground truth

latent dim = 1024

Visualizing latent space

latent dim = 1024 (visualizing by t-sne)