psd's Introduction

Unsupervised Discovery of Parts, Structure, and Dynamics

This repository contains pre-trained models and testing code for PSD presented at ICLR 2019.

http://psd.csail.mit.edu

Installation

Our current release has been tested on Ubuntu 18.04 LTS.

Cloning the repository

git clone [email protected]:zhenjia-xu/psd.git
cd psd

Set up Python environment:

pipenv install --dev

Guide

Generating shape dataset

pipenv run python data_generator.py

Downloading the pretrained model (185 MB)

./download_model.sh

Visualization

pipenv run python demo.py --gpu ID

--gpu ID: use which gpu (starting from 0). Set to -1 to use CPU only.

The results will be presented in ./demo/index.html.

Reference

@inproceedings{psd,
  title={Unsupervised Discovery of Parts, Structure, and Dynamics},
  author={Xu, Zhenjia and Liu, Zhijian and Sun, Chen and Murphy, Kevin and
        Freeman, William T and Tenenbaum, Joshua B and Wu, Jiajun},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2019},
}

For any questions, please contact Zhenjia Xu ([email protected]).

psd's People

Contributors

Stargazers

Watchers

psd's Issues

Could you please provided the pre-trained model on exercise dataset?

Thanks for your help !

Recognizing of object parts and its hierarchical structure

Hello
How are you?
Thanks for contributing this paper and project.
I have a question.
It looks likely that this paper recognizes the parts and hierarchical structure of ONLY object (eg: running person) whose part has local movement.
For example, does this system support to recognize the parts and its hierarchical structure of a running car?
Thanks