FHAG-with-BOIL

Overview

This is the implementation of the method proposed in "Distortion Model based Spectral Augmentation for Generalized Recaptured Document Detection" with pytorch(1.9.0, gpu version). The associated datasets are available upon request.

Introduction

In this work, we improve DPAD approaches by addressing two limitations of existing FDA methods. We establish a BOIL method that locates the BOI related to the recapturing operation and propose a FHAG strategy that enhances halftoning features in the BOI.

Environment Request

python == 3.10.6
fire == 0.4.0
matplotlib == 3.3.4
numpy == 1.20.2
opencv_python == 4.5.2.54
Pillow== 9.4.0
scikit_learn == 1.2.1
torch == 1.9.0
torchnet == 0.0.4
torchvision == 0.10.0
tqdm == 4.61.1

Files structure

• data
• models
• utils
• config.py
• main.py
• FHAG.py
• BOIL.py
• analyze.py

Files Description

BOIL.py
Calculate the BOI of existing sample

FHAG.py
Augment the recapturing features by amplitude manipulation

Run

Train

python main.py train 

Test

python test.py 

Analyze

python analyze.py

Citation

If you use our code please cite: Changsheng Chen, Bokang Li, Rizhao Cai, Jishen Zeng, and Jiwu Huang, Distortion Model based Spectral Augmentation for Generalized Recaptured Document Detection, in IEEE Transactions on Information Forensics and Security, 2023

The datasets presented in our work are available upon request.