Detecting-Generated-Images-by-Real-Images

Setup

python >= 3.8 torch >= 1.8.0

Make LNP dataset

We have extracted the LNP using the code in CycleISP, replacing the corresponding code in CycleISP with our provided dataset_rgb.py, denoising_rgb.py and test_sidd_rgb.py. If you want to generate real images LNP and fake images LNP, please swap the positions of 0_real and 1_fake in dataset_rgb.py.

python test_sidd_rgb.py

Classification

Take the produced LNP dataset and sort it using the code provided by CNNDetection. Use resnet.py to replace the code in it.

python train.py --name model --dataroot ./dataset/ --classes airplane,bird,bicycle,boat,bottle,bus,car,cat,cow,chair,diningtable,dog,person,pottedplant,motorbike,tvmonitor,train,sheep,sofa,horse

Acknowledgments

This repository borrows content from CycleISP, as well as the CNNDetection repository.