Deepfake Detection

Overview

This project aims to classify videos as "REAL" or "FAKE" using a deep learning model. The model architecture leverages the InceptionV3 model for feature extraction, followed by LSTM and GRU layers for sequence processing.

Getting Started

Clone the Repository:

git clone https://github.com/your-repo-link.git
cd your-repo-directory

Install Dependencies:

pip install -r requirements.txt

Run the Application:

python app.py

Dataset

The dataset used for training and evaluation consists of:

• 77 Fake Videos
• 76 Real Videos

Each video has 10 frames extracted based on motion detection.

This newly motion detected dataset is released on kaggel as Deepfake Detection Challenge Dataset - Face Images.

Architecture

The model architecture consists of the following components:

1. Feature Extraction:

   • InceptionV3 pre-trained on ImageNet is used for feature extraction.
   • GlobalAveragePooling2D is applied to pool features from each frame.

2. Sequential Processing:

   • TimeDistributed applies the feature extractor to each frame independently.
   • LSTM and GRU layers are used to capture temporal dependencies.
   • Dropout is applied for regularization to prevent overfitting.
   • Dense layers are used for the final classification.

Training

The model was trained on Kaggle Notebooks with the following specifications:

• GPU Accelerators: T4 x 2
• Dataset: Deepfake Detection Challenge Dataset.

Demo

You can try out the model live on Hugging Face Spaces: Demo

Contributing

Contributions are welcome! If you'd like to contribute, please fork the repository and create a pull request with your changes. Ensure your code adheres to the existing style and includes appropriate tests.

License

This project is licensed under the MIT License - see the LICENSE file for details.