dvpinho / RealFakeFaces

Classification of real-fake faces using neural networks.

Real Fake Face Image Classifier "DeepFakes"

This repository contains the project notebook that was developed during the Module #3 Deep learning, computer vision and evolutionary computation course by TDX | University of Coimbra Data Science Academy (Powered by Talkdesk). Total course time was 80 hours plus 3 additional weeks for project related work.

For the project work and with the provided dataset, a real-fake face classifier was developed using several approaches: convolutional neural networks (CNN) and transfer learning with and without face cropping using the Multi-Task Cascaded Convolutional Neural Networks (MTCNN).

To extend the course work, and in order to leverage the created models, a web application was developed using the streamlit framework. For more information about the models or the project, check out the project folder in the root directory of this repository. The application source code is in the app folder.

Run the app locally

Open a terminal and start by cloning the project repository

  git clone https://github.com/dvpinho/RealFakeFaces.git

Go to the project directory

  cd RealFakeFaces

Install dependencies

  pip install -r requirements.txt

Start the server (Linux / MacOS)

  streamlit run app/app.py

Start the server (Windows)

  streamlit run app\app.py

If the browser window does not open automatically when the streamlit run command is executed, you can manually navigate to localhost:8501 To terminate the application, go back to the terminal and shutdown the server by pressing CTRL + C.

License

MIT

Screenshots