REAL-TIME-DEEPFAKE-DETECTION-FOR-ONLINE-MEDIA-INTEGRITY

This project involves the development and implementation of a deep learning algorithm that leverages a CNN-RNN approach to detect deepfake videos. This solution is designed to work across various online platforms.

Table of Contents

• Introduction
• Features
• Technologies Used
• Model Architecture
• Installation
• Results
• Model
• Dataset
• Contributing
• Contact

Introduction

Deepfake technology has made it increasingly difficult to distinguish between real and manipulated media. This project aims to address this challenge by developing a robust deep learning model that can accurately identify deepfake videos. By utilizing a combination of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), the model achieves high accuracy and low loss in detecting manipulated media.

Features

• High Accuracy: Achieved an accuracy of 86% in detecting deepfake videos.
• Low Loss: Model achieved a loss of 0.19, indicating robust performance.
• CNN-RNN Approach: Utilizes the strengths of both CNNs and RNNs for feature extraction and temporal analysis.

Technologies Used

• Python: Main programming language.
• TensorFlow/Keras: Deep learning framework for building and training the model.
• OpenCV: For video processing.
• NumPy: For numerical computations.
• Pandas: For data manipulation and analysis.
• Tkinter: For creating the graphical user interface.
• Google Colab: For running and sharing Jupyter notebooks.
• pyttsx3: For text-to-speech conversion.

Model Architecture

Installation

1. Clone git repo

$ git clone "https://github.com/PRADULOP/REAL-TIME-DEEPFAKE-DETECTION-FOR-ONLINE-MEDIA-INTEGRITY.git"

2. Create virtual env

$ pip install virtualenv

$  python -m venv env

$  ./env/Scripts/activate

3. Import Modules

$ pip install keras

$ pip install pillow

$ pip install numpy

$ pip install tensorflow

$ pip install opencv-python     

$ pip install tkinter     

$ pip install cvzone  

$ pip install pyttsx3  

$ pip install torchvision

4. Execute prediction model

$ python ./main.py

Results

The model achieved the following performance metrics:

• Accuracy: 86%
• Loss: 0.19

These results demonstrate the model's robust performance in identifying deepfake videos.

Model

Model Download Link: https://drive.google.com/file/d/1cFnH7LE5qiS6WaShWLSo58B-1UgrH4AS/view?usp=sharing

Dataset

This model was trained using an aggregated dataset consisting of:

• Celeb-DF: A large-scale deepfake dataset containing celebrity videos.
• DFDC: Deepfake Detection Challenge dataset provided by Facebook.
• FaceForensics++: A dataset of manipulated videos for deepfake detection.

Aggregated Dataset Download Link: https://drive.google.com/drive/folders/1umdXB_4gLBHH6LoEyi5X7E9rwm6on1YQ

Contributing

Contributions are welcome! Please feel free to submit a Pull Request or open an Issue to discuss potential changes.

Contact

For any questions or inquiries, please contact me at [email protected].