GRASS: Generative Recursive Autoencoders for Shape Structures

By Jun Li, Kai Xu, Siddhartha Chaudhuri, Ersin Yumer, Hao Zhang, Leonidas Guibas

This repository contains the pre-trained models for box structure generation, as well as the training/testing code for the generation model.

Details of the work can be found here.

A PyTorch implementation (currently with only the VAE part) is available at: https://github.com/kevin-kaixu/grass_pytorch.

Citation

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

@article {li_sig17,
    title = {GRASS: Generative Recursive Autoencoders for Shape Structures},
    author = {Jun Li and Kai Xu and Siddhartha Chaudhuri and Ersin Yumer and Hao Zhang and Leonidas Guibas},
    journal = {ACM Transactions on Graphics (Proc. of SIGGRAPH 2017)},
    volume = {36},
    number = {4},
    pages = {to appear},
    year = {2017}
}

Guide:

Training

Run trainTestVaeGan.m to train the vae-gan model on the provided chair dataset.

Testing

Use test_demo.m to generate shapes based on trained model. There is already a pre-trained model inside. The generated shape structures could be visulized in matlab.

For any questions, please contact Jun Li([email protected]) and Kai Xu([email protected]).