sutiankang/pytorch-lightning-pointnet

pytorch-lightning-pointnet

This repo is implementation for PointNet in pytorch-lighting framework.

Prerequisites

• Python 3.10.6

conda create -n pl_pointnet python=3.10.6

• Pytorch 2.0.1+cu118 Torchvision 0.15.2+cu118 Torchaudio 2.0.2+cu118

  Install pytorch by conda as:

conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.8 -c pytorch -c nvidia 

or pip as:

pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118

• Install requirements

pip install -r requirements.txt

Data Preparation(ModelNet10/Modelnet40)

Download

The modelnet10/40 dataset can be downloaded from here.

Train

python train.py

Hyper-parameter Tutorial

• The common setting (including ddp, early stopping or optimizer and any other tricks) can be set from here.
• If you want to speed up dataset loading using offline data, you can set the parameter use_cache to True from here.
• If you want to get better performance by adding normals, you can set the paramater use_normals to True from here.

Results

Model	Accuracy
PointNet (Official)	89.2
Pointnet_Pointnet2_pytorch (Pytorch with normal)	91.4
pytorch-lightning-pointnet (pytorch-lightning with normal)	91.7

Weights

You can download our weight after training from here.

Reference

Pytorch_Pointnet_Pointnet2

Citation

If you find this repo useful in your research, please consider citing it and our other works:

@article{pytorch-lightning-pointnet,
      Author = {Tiankang Su},
      Title = {pytorch-lightning-pointnet},
      Journal = {https://github.com/sutiankang/pytorch-lightning-pointnet},
      Year = {2024}
}