1st Solution of MMDD 2025: The 1st Multimodal Deception Detection Competition @MM2025

LCUNet: A Lightweight Concatenated Unified Mapping Multi-modal Deception Detector (^▽^)

This is official Pytorch implementation of "LCUNet: A Lightweight Concatenated Unified Mapping Multi-modal Deception Detector"

Team

Team: WHU_PB
Team leader: Glenn_xxy (Real name: Xiang Xinyu)

Before running:

Place the training and test datasets inside the ./dataset2/
put our ckpt is in ./checkpoint/

Train:

sh train_test_feature_phase2_upload.sh

Test:

python test_phase2_final_upload.py

Environment

Please check the environment.yml for details

Checkpoint

https://pan.baidu.com/s/1Xr5-DGyH48NhR0xg64R2zQ?pwd=uiaf
pw: uiaf

Ours result:

Please check the ./result

More details:

If there are any problems with the operation, please feel free to contact me at any time. I will be happy to answer any questions you have.

Email me:
xiangxinyu@whu.edu.cn 
1977587176@qq.com

Thanks:

@inproceedings{guo2023audio,
  title={Audio-visual deception detection: Dolos dataset and parameter-efficient crossmodal learning},
  author={Guo, Xiaobao and Selvaraj, Nithish Muthuchamy and Yu, Zitong and Kong, Adams Wai-Kin and Shen, Bingquan and Kot, Alex},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={22135--22145},
  year={2023}
}
@article{guo2024benchmarking,
  title={Benchmarking Cross-Domain Audio-Visual Deception Detection},
  author={Guo, Xiaobao and Yu, Zitong and Selvaraj, Nithish Muthuchamy and Shen, Bingquan and Kong, Adams Wai-Kin and Kot, Alex C},
  journal={arXiv preprint arXiv:2405.06995},
  year={2024}
}