aadd-2025

This repository contains a dataset of deepfake images and a classifier for the “Deepfake Detection Challenge 2025”.

Dataset Description.

The dataset includes thousands of deepfake images generated using advanced techniques (e.g., GAN). The images represent human faces with synthetic alterations, useful for testing detection algorithms.

• Image number: ~50,000 (example, edit with your own number).
• Format: JPG/PNG.
• Size: 30 GB (full dataset).

Challenge.

“Deepfake Detection Challenge 2025” invites researchers to develop classifiers to distinguish real images from deepfakes. Use the dataset to train and test your model!

Contents.

• sample_data/: 10 sample images (5 MB).
• code/: Classifier code (classifier.py).
• Full dataset: Available on OneDrive.

Download.

• Complete Dataset: Download from OneDrive.
• Instructions:
  1. Download Deepfake_Dataset.zip (30 GB).
  2. Extract to the root of the repo.

Usage.

1. Download and extract the dataset from OneDrive.
2. Install the dependencies: pip install -r requirements.txt.
3. Run the classifier: python code/classifier.py.

License.

The dataset and code are under CC BY-NC 4.0 with restriction:

• Attribution: You must cite “Mirko Rossi” and the paper “Deepfake Detection, Rossi et al., 2025.”
• Noncommercial: Commercial use prohibited.
• Challenge only: The dataset can only be used for the “Deepfake Detection Challenge 2025.”

Citation.

• Paper: “Deepfake Detection, Rossi et al., 2025” (update with your title).
• Dataset: [INSERT-THEN-LINK]