SACRW

This repository contains the official implementation of the NeurIPS 2023 paper:

Object-centric Learning with Cyclic Walks between Parts and Whole

Ziyu Wang, Mike Zheng Shou, Mengmi Zhang

Access to our poster HERE and presentation video HERE.

Environment Setup

The basic environment contains these packages:

• Python 3.9.17
• torch 1.12.1
• torchvision 0.13.1
• pillow 9.2.0
• pytorch-lightning 1.1.4
• opencv-python 4.8.0.74

Other dependencies can be installed as needed.

Dataset

For PascalVOC 2012 and COCO 2017, download the datasets from the following links:

• PascalVOC 2012
• COCO 2017

For MOVi-C and MOVi-E, please follow this repository.

For Birds, Dogs, Cars, and Flowers, please follow this repository.

Training & Testing

To train the model from scratch, please follow the steps below:

• Modify the data_paths in train.py to your own.
• Run the command as shown in the following example. The input parameter is the ID of your GPUs.

sh script/train_sacrw_voc.py 0,1,2,3

To test the model:

• Modify the data_paths in test.py to your own.
• Run the command as shown in the following example. The input parameter is the ID of your GPUs.

sh script/test_sacrw_voc.py 0,1,2,3

Citation

If you find our paper and/or code helpful, please cite:

@article{wang2023object,
  title={Object-centric Learning with Cyclic Walks between Parts and Whole},
  author={Wang, Ziyu and Shou, Mike Zheng and Zhang, Mengmi},
  journal={arXiv preprint arXiv:2302.08023},
  year={2023}
}