CFL-Net: Image Forgery Localization Using Contrastive Learning

This is the official implementation of WACV-2023 paper by Fahim Faisal Niloy,Kishor Kumar Bhaumik,and Simon S. Woo
CFL-Net: Image Forgery Localization Using Contrastive Learning.

Setup: Run

pip install -r requirements.txt

Step 1: Download IMD-20 Real Life Manipulated Images from Link.

step2: set the dataset path in

for example, if you have downloaded and unzipped the IMD2020 dataset in the following directory: /home/forgery/ then put /home/forgery/ as the base_dir in the config file. (DO NOT put /home/forgery/IMD2020/ in base_dir )

Step 3: To train the model run

python trainer.py

Step 4: To test the model run

python evaluate.py

Please Cite Our paper using:

@inproceedings{niloy2023cfl,
title={CFL-Net: image forgery localization using contrastive learning},
author={Niloy, Fahim Faisal and Bhaumik, Kishor Kumar and Woo, Simon S},
booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
pages={4642--4651},
year={2023}}