HAMI-M3D

This repo is the released code of our work Harmfully Manipulated Images Matter in Multimodal Misinformation Detection

📣 News: This paper is accepted by ACM MM'24!

Our released code follows to "EANN: Event Adversarial Neural Networks for Multi-Modal Fake News Detection" and "BDANN: BERT-Based Domain Adaptation Neural Network for Multi-Modal Fake News Detection"

Requirements

torch==1.12.1
cudatoolkit==11.3.1
transformers==4.27.4

Train

• Prepare the datasets Weibo, Gossip and Twitter. The datasets are from https://github.com/yaqingwang/EANN-KDD18 and https://github.com/shiivangii/SpotFakePlus, and you should put them in ./Data

• Run the python file

cd src
python ./run.py

• Check log files in ./log

Tips

1. You should manually split the training set of GossipCop into the divisions of training and validation, then, revise the file road in the function write_data in line 89, process_gossipcop.py
2. We prepare a auto_logging.py to automatically read the output log files into an excel table.

Citation

@inproceedings{wang2024harmfully,
  author       = {Bing Wang and
                  Shengsheng Wang and
                  Changchun Li and
                  Renchu Guan and
                  Ximing Li},
  title        = {Harmfully Manipulated Images Matter in Multimodal Misinformation Detection},
  booktitle    = {{ACM} International Conference on Multimedia},
  pages        = {2262--2271},
  publisher    = {{ACM}},
  year         = {2024},
  doi          = {10.1145/3664647.3681322},
}