CVmeta / recaptured-image-detection-ICDIP-2022

recaptured-image-detection-ICDIP-2022

An open source code for paper "Identifying Recaptured Images Using Deep Correlation"

Our Environemnt Setting

• Ubuntu 20.04
• Python 3.8
• NVIDIA GPU CUDA 10.2 + cuDNN 8.1.0
• PyTorch 1.9.1
• Pillow 8.3.1

Data Path

• Train: ./data/train_image/NI ./data/train_image/RI
• validation: ./data/val_image/NI ./data/val_image/RI
• Test: ./data/test_image/NI ./data/test_image/RI

Run

• Training: Set parameters in 'train.py' or at the command line.

python train.py --batch-size 64 --patch-size 96 --epoch 400 --lr 0.001

• Testing: Run python test.py.

Acknowledgment

This code refers to JCST_NIvsCG.