detector_codes/DFFreq-main/README.md

Dual Frequency Branch Framework with Reconstructed Sliding Windows Attention for AI-Generated Image Detection

Jiazhen Yan¹, Ziqiang Li¹, Fan Wang¹, Ziwen He¹, Zhangjie Fu^1‡

^‡Corresponding author

¹Nanjing University of Information Science and Technology

🔥 News

• [2026-02-07]🎉🎉🎉 DFFreq is accepted by IEEE Transactions on Information Forensics & Security.

⏳ Quick Start

1. Installation

conda create -n DFFreq -y python=3.9
conda activate DFFreq
pip3 install torch torchvision
pip install -r requirements.txt 

2.Getting datasets

Datasets	Paper	Url
GANGen-Detection	Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning (AAAI 2024)	Google Drive
DiffusionForensics	DIRE for Diffusion-Generated Image Detection (ICCV 2023)	Google Drive
UniversalFakeDetect	Towards Universal Fake Image Detectors that Generalize Across Generative Models (CVPR 2023)	Google Drive
AIGCDetectBench	PatchCraft: Exploring Texture Patch for Efficient AI-generated Image Detection	ModelScope
AIGIBench	Is Artificial Intelligence Generated Image Detection a Solved Problem? (NeurIPS 2025)	Huggingface/Baidu Netdisk

3.Inference

Of course, you need to change [DetectionTests] in test.py when testing.

We present our inference results in log_test.log.

python test.py --model_path ./checkpoints/model_epoch_last.pth

⏳ Training

The training set uses four classes from CNN-Spot(CNN-generated images are surprisingly easy to spot...for now, CVPR 2020): car, cat, chair, and horse. Baidu Netdisk

python train.py --name 4class-car-cat-chair-horse --dataroot [training datasets path] --classes car,cat,chair,horse

Citation

@article{yan2026dual,
  title={Dual Frequency Branch Framework with Reconstructed Sliding Windows Attention for AI-Generated Image Detection},
  author={Yan, Jiazhen and Li, Ziqiang and Wang, Fan and He, Ziwen and Fu, Zhangjie},
  journal={IEEE Transactions on Information Forensics and Security},
  year={2026},
  publisher={IEEE}
}

Contact

If you have any question about this project, please feel free to contact 247918horizon@gmail.com