This Deep Fake Detection system is designed to identify manipulated media, focusing on images, using advanced machine learning techniques. It employs TensorFlow and PyTorch frameworks to analyze visual content and distinguish between genuine and altered media.
- Utilization of TensorFlow and PyTorch for deep learning model development.
- Implementation of Convolutional Neural Networks (CNNs) and Transformer model.
- Data preprocessing and augmentation to enhance the model's ability to generalize.
Ensure you have Python installed on your system and follow these steps:
- Clone the repository
- Navigate to the project directory
- Install required dependencies: pip install -r requirements.txt
The project uses the "Deepfake and Real Images" dataset from Kaggle, consisting of real and deepfake-generated images.
Source: Sourced from Kaggle [https://www.kaggle.com/datasets/manjilkarki/deepfake-and-real-images], provided by user Manjil Karki, which is a part of the OpenForensics dataset [https://zenodo.org/records/5528418#.YpdlS2hBzDd]. This dataset is specifically designed for the challenging task of multi-face forgery detection and segmentation in-the-wild.
The OpenForensics dataset used in this project was introduced in the following research paper:
@inproceedings{ltnghia-ICCV2021,
title = {OpenForensics: Large-Scale Challenging Dataset For Multi-Face Forgery Detection And Segmentation In-The-Wild},
author = {Trung-Nghia Le and Huy H. Nguyen and Junichi Yamagishi and Isao Echizen},
booktitle = {International Conference on Computer Vision},
year = {2021},
}