xingyu-liu / SP-GAN

SP-GAN: Sphere-Guided 3D Shape Generation and Manipulation (SIGGRAPH 2021)

This repository contains a Pytorch implementation of the paper:

SP-GAN: Sphere-Guided 3D Shape Generation and Manipulation.
Ruihui Li, Xianzhi Li, Ka-Hei Hui, Chi-Wing Fu.

Dependencies

• Python 3.6
• CUDA 10.0.
• PyTorch. Codes are tested with version 1.2.0
• (Optional) TensorboardX for visualization of the training process.

Following is the suggested way to install these dependencies:

# Create a new conda environment
conda create -n PointAugment python=3.6
conda activate PointAugment

# Install pytorch (please refer to the commend in the official website)
conda install pytorch=1.2.0 torchvision cudatoolkit=10.0 -c pytorch -y

Usage

As a generation task, you can freely customize your own dataset, such as ShapeNet , SMPL, SMAL

Or directly download our prepared H5 Files from here.

To train a model on point clouds sampled from 3D shapes:

python train.py --choice Chair

Log files and network parameters will be saved to log folder in default.

For your reference, the pretrained models are also available from here

We provide various visulization function for shape interpolation, part interpolation, and so on.

python visual.py

Evaluation

Please ref to Latent-GAN, PointFlow, and PDGN

Citation

If you find our work useful in your research, please consider citing:

@article{li2021spgan,
  title={{SP-GAN}:Sphere-Guided 3D Shape Generation and Manipulation},
  author={Li, Ruihui and Li, Xianzhi and Hui, Ke-Hei and Fu, Chi-Wing},
  journal={ACM Transactions on Graphics (Proc. SIGGRAPH)},
  volume={40},
  number={4},
  year={2021},
  publisher={ACM}
}