Official repository for the WACV 2024 paper Multi-view Classification with Hybrid Fusion and Mutual Distilation. Here, you'll find our code to train and evaluate our method, MV-HFMD. Currently, we provide code to run MV-HFMD on the Hotels-8k dataset.

Instructions

To train our method on Hotels-8k, first, download the dataset from this link. Unzip the file into the desired directory. Then, run

python3 main.py --data-directory {DATA_DIRECTORY}

You can toggle the mutual distillation loss function with the argument

--use_mutual_distillation_loss {True/False}

And then the number of images per collection that you wish to train and evaluate on

--num_images {2/3/4}

By default, the model will generate classification predictions for each individual image and then the entire multi-view collection. These are given in the model output dictionary under the keys 'single' and 'mv_collection', respectivefully.

Requirements:

• Python 3
• torch
• numpy
• timm
• einops

Citation:

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

@inproceedings{black2024multi,
  title={Multi-View Classification Using Hybrid Fusion and Mutual Distillation},
  author={Black, Samuel and Souvenir, Richard},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={270--280},
  year={2024}
}