A Lightweight and Effective Image Tampering Localization Network with Vision Mamba

An official implementation code for paper "A Lightweight and Effective Image Tampering Localization Network with Vision Mamba". This repo provides code and trained weights.

Framework

Dependency

• torch 1.13.1+cu117
• torchvision 0.14.1+cu117
• python 3.10
• causal-conv1d 1.0.0
• mamba-ssm 1.0.1
• selective_scan 0.0.2

Usage

For example to train:

python generate_npy.py
python train.py 

For example to test: download ForMa_weights.pth form Google Drive or Baidu Netdisk or 123 Pan and place it in the weights/ directory.

python generate_npy.py
python test.py 

Citation

If you use this code for your research, please cite our paper

@article{guo2025light,
  title={A Lightweight and Effective Image Tampering Localization Network with Vision Mamba},
  author={Guo, Kun and Cao Gang and Lou, Zijie and Huang, Xianglin and Liu, Jiaoyun},
  journal={IEEE Signal Processing Letters},
  year={2025}
}