2024-MM-VITAL

[ACM MM 2024] Code for the paper "Robust Variational Contrastive Learning for Partially View-unaligned Clustering"

Installation

1. Clone the repository:

   git clone https://github.com/He-Changhao/2024-MM-VITAL

2. Install the Python dependencies:

   numpy>=1.26.3
   scipy>=1.11.4
   scikit-learn>=1.4.0
   munkres>=1.1.4
   torch>=1.12.1
   matplotlib>=3.8.2
   pyyaml>=6.0.1
   

Dataset and Configuration

The datasets used in our paper can be downloaded from Google Drive. Each dataset's configuration is written in a .yaml file located in the config folder. The parameters are explained in the get_args_parser() function in run.py. The structure of the datasets and their corresponding .yaml files should be as follows:

VITAL-path
    └─── datasets
        │   CUB.mat
        │   Deep Animal.mat
        │   Deep Caltech-101.mat
        │   MNIST-USPS.mat
        │   NoisyMNIST.mat
        │   NUS-WIDE.mat
        │   Scene-15.mat
        │   WIKI.mat
    └─── config
        │   CUB.yaml
        │   Deep Animal.yaml
        │   Deep Caltech-101.yaml
        │   MNIST-USPS.yaml
        │   NoisyMNIST.yaml
        │   NUS-WIDE.yaml
        │   Scene-15.yaml
        │   WIKI.yaml

If you need to add new datasets for training, please modify dataloader.py according to the existing dataset format and add the corresponding .yaml file to the config directory.

Usage

All dataset parameters can be set by modifying the ./config/*.yaml files. After editing the configuration file, you can train the model using the following command:

python vital_path/run.py --dataset_name 'CUB'

Experiment Results

The partially aligned (50%) clustering performance:

The fully aligned (100%) clustering performance:

Citation

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

@inproceedings{he2024robust,
  title={Robust Variational Contrastive Learning for Partially View-unaligned Clustering},
  author={He, Changhao and Zhu, Hongyuan and Hu, Peng and Peng, Xi},
  booktitle={ACM Multimedia 2024},
  year={2024}
}