Deepfake Detector
Deepfake Detector
A deepfake detector built using NextJS, FastAPI and Tensorflow
View Demo
Table Of Contents
About The Project
Deepfake Detector is an AI course final project that consists of building a deep learning model that can detect deepfake photos by using AI techniques. The model was built using a Convolutional Neural Network (CNN) and trained using the OpenForensics dataset. The dataset contains aproximately 191,000 images. The training set contains 140,000, the validation set contains 40,000 and the test set contains 11,000 images. The images are divided into two classes: real and fake.
The model was built in Google Colab Pro with Tensorflow and Keras API. The trained model was saved as a .tf file and then used in a web service built with FastAPI. The web service was deployed to Render and the client was built using NextJS and deployed to Vercel.
The model was trained using 20 epochs and achieved an accuracy of 0.91 and a loss of 0.2439 in the test set.
Built With
The technologies used for this project were:
- NextJS for the client
- FastAPI for the server
- Tensorflow for the model
Getting Started
To use this application, simply clone the repository, move into the client and server directories, install the dependencies, and start the development server. Here are the steps:
- Clone the repo
git clone https://github.com/davidperjac/react-wordle-clon.git
Client Installation
- Move into the client directory
cd client
- Install NPM packages
npm install
- Start the development server
npm run dev
- Open your web browser and go to
http://localhost:3000
Server Installation
- Move into the server directory
cd server
- Create a virtual environment
python3 -m venv venv
- Activate the virtual environment
source venv/bin/activate
- Install the requirements
pip install -r requirements.txt
- Start the development server
uvicorn main:app --reload
Usage
Landing Page
Model Page
Code Page
Authors
Built with ❤️ by:
- David Perez - David Perez
- Fernando Bucheli - Fernando Bucheli
- David Bravo - David Bravo
