Data-Independent Operator: A Training-Free Artifact Representation Extractor for Generalizable Deepfake Detection

Chuangchuang Tan, Yao Zhao, Ping Liu, Guanghua Gu, ShiKui Wei, Baoyuan Wu, Yunchao Wei

Beijing Jiaotong University, YanShan University, A*Star, CUHK

Reference github repository for the paper Data-Independent Operator: A Training-Free Artifact Representation Extractor for Generalizable Deepfake Detection.

⭐ If Our code is helpful to your images or projects, please help star this repo. Thanks! 🤗

Overall Pipeline

TODO

~~[ ] Update link to paper~~
Pretrained models
Dataset download
Code released

Detection Results on Various Generators

Results on ForenSynths Dataset.

Results on GANGen-Detection Dataset. download.

Results on DiffusionForensics Dataset.

Results on UniversalFakeDetect Dataset.

Results on AIGCDetectBenchmark Dataset.

Environment setup

scipy
scikit-learn
numpy
opencv_python
Pillow
torch>=1.2.0
torchvision
gdown

pip install -r requirements.txt

Prepare Dataset

chmod 777 ./download_dataset.sh
sh ./download_dataset.sh

Testing the detector

wget https://lid-1302259812.cos.ap-nanjing.myqcloud.com/tmp/MDIO2github.pth -O MDIO.pth
CUDA_VISIBLE_DEVICES=0 python test.py --model_path ./MDIO.pth --batch_size {BS}

Acknowledgments

This repository borrows partially from the CNNDetection.

chuangchuangtan/Data-Independent-Operator