Employing Reinforcement Learning to Construct a Decision-making Environment for Image Forgery Localization

Table of Contents

• Introduction
• Dependency
• Installation
• Usage

Introduction

The widespread misuse of advanced image editing tools and deep generative techniques has led to a proliferation of images with altered content in real-life scenarios, often without any discernible traces of tampering. This has created a potential threat to security and credibility of images. Image forgery localization is an urgent technique. In this paper, we propose a novel reinforcement learning-based framework CoDE (Construct Decision-making Environment) that can provide reliable localization result of tampered area in forged images. We model the forgery localization task as a Markov Decision Process (MDP), where each pixel is equipped with an agent that performs Gaussian distribution-based continuous action to iteratively update the respective forgery probability, so as to achieve pixel-level image forgery localization.

The proposed Markov decision process for image forgery localization.

The proposed CoDE for image forgery localization task. FPM denotes the forgery probability map.

Visual results for image sforgery localization on different benchmark datasets using CoDE.

Dependency

• python 3.7.0
• torch 1.7.1+cu110

Installation

You can run the following commands to create a new environment:

conda create -n env python=3.7.0
pip install -e requirement.txt

Or you can directly run this command to setup your environment:

conda env create -f environment.yaml

Usage

Data Preparation:

Put the images in "dataset/forgery/" and the ground-truth of masks in "dataset/gt/", then generate the index "data.txt" under "dataset/".
We have given the template of "dataset/", please check in the directory for more details.

Pre-trained Model:

The pre-trained model can be downloaded from: Google Drive or Baidu Pan (Code:CoDE)
The downloaded file should be placed in "weights/".

Training Command:

python train.py --DATASET dataset --RATIO 0.7 --TRAIN-BATCH-SIZE 8

Here we only present the basic training setting, you can browse 'train.py' for more details.

Testing Command:

python test.py --WEIGHT-PATH ./weights/pretrained_CoDE.pth --DATASET dataset --RATIO 0 --VISUALIZE 0

When --VISUALIZE is True (1), the results for action map, forgery probability map and predicted mask will be saved in the 'result/'.