PointNet 3D Conversion & Reconstruction

This repository is a modularized version of the Colab project point_net.py. It contains implementations for PointNet-based classification and an autoencoder for point cloud reconstruction (using Chamfer Distance).

Structure

pointnet-3d-conversion/
├── README.md
├── requirements.txt
├── LICENSE
├── .gitignore
├── data/
│   └── README.md
└── src/
    ├── dataset.py
    ├── model.py
    ├── utils.py
    ├── train_classify.py
    ├── train_reconstruct.py
    └── __init__.py
└── main.py

Quick start

1. Install dependencies:

pip install -r requirements.txt

2. Place ModelNet40 extracted folder modelnet40-normal_numpy next to this repo or let the script download it.
3. Run the full pipeline (classification + reconstruction):

python main.py --data_dir ./modelnet40-normal_numpy --epochs 10 --batch_size 64

Notes

• The dataset is large; do not push it to GitHub. Keep data/README.md as instruction.
• This repo was generated from a Colab-exported file by Shubham Raj.