/hackprix-codement

Deepfake Detector build using ResNext and LSTM at HACKPRIX hackathon

Primary LanguageHTML

Deep fake detection Django Application

Deepfake detector is a tool capable of detecting facially altered videos using ResNext and LSTM networks.

  • The model is trained using over 700 videos and we achieved a prediction accuracy of 90%.
  • We also created a user-friendly web application using Django allowing users to upload videos for deepfake detection.

Requirements:

  • CUDA version >= 10.0 for GPU

  • GPU Compute Capability > 3.0

  • Required modules are listed in the requirements.txt

Python >= v3.6
Django >= v3.0

How to run the application on local machine

step 1: Download and Install Cuda version 12.1 (most compatible)

download link: https://developer.nvidia.com/cuda-12-1-0-download-archive

Step 2: Install requirements.txt

Step 3: open folder Deep_fake_Django & start the server

python manage.py runserver

step 4: Go to localhost 127.0.0:8000

step 5: Upload the video to be tested and select the bar to 40 frames. (Test videos are listed in folder - Test_Videos)