This repository contains PyTorch implementation of Learning Representations and Generative Models for 3D Point Clouds by Panos et. al. The aim of the paper is to analyse different generative models.
Notebook main.ipynb contains code to all the training experiments along with interactive 3D point cloud representation. For ease, access the Table of contents.
The repository includes implementation of:
- 3D interactive plot
- Chamfer Loss and Earth Mover Distance
- Autoencoder for 3D point cloud
- raw-GAN
- l-GAN
For Earth Mover Distance I have used an iterative algorithm to calculate matching matrix
Note : Models are not trained because of limited resources. If anyone can help, I would be highly obliged.