SJSU_Cmpe195F - Fall 2021 - Deepfake Detection Project

Github Repo

https://github.com/ThaiQ/Deepfake_Recognition

Demo

https://www.youtube.com/watch?v=ga2ilTjN0pk&ab_channel=David

Dependencies

1. NodeJS: https://nodejs.org/en/download/
2. Python3: https://www.python.org/downloads/
3. Tensorflow 2.0 or Cuda if GPU (Optional)

Description

This is a student project that aims to build a deepfake detection web application.

High level Architecture: React-Flask stack

Design Diagram

CNN Architecture

Server installation

1. CD to server
2. pip install -r ./Deepfake_Detection_NN/requirement.txt
3. Download our ensemble models
4. Save downloaded models in server/models - Make sure its .h5 files - script will handle parsing.

5. Run server python app.py

Frontend installation

1. CD to app
2. npm install - install node dependencies
3. Run react app npm start

If you want to train your own model:

1. Follow server installation to step 2
2. Get data at https://www.kaggle.com/c/deepfake-detection-challenge/data
3. run train.py in ./server/Deepfake_Detection_NN

Sample Model Output

Where fakes are 50% and greater.