DeepShield - DeepFake Detection

DeepFake Detection Web-App 🖥 using Deep Learning (ResNext and LSTM), Flask, and ReactJs where you can predict whether a video is FAKE or REAL along with the confidence ratio.

Explanation of the Project

• Designed a DeepFake Detection system to identify DeepFake videos using Deep Learning techniques like ResNext and LSTM. Integrated the trained model with a ReactJS Frontend and Flask Backend.
• The dataset used to train the model is available here.
• Trained model files can be found here.
• Model training reference taken from this resource.

Project Structure

DeepShield
    |
    |--- DeepFake_Detection
    |--- Implementation Video
    |--- Project-Setup.txt
    |--- Requirements.txt

1. DeepFake_Detection - Root folder of the project.
2. Implementation Video - Shows complete working of the project.
3. Project-Setup.txt - Contains all steps to set up and run the project.
4. Requirements.txt - Python libraries needed for this project.

Project Set-up Guidelines

Setup instructions are listed here.

Notes

1. In the root folder (DeepFake_Detection), create a folder called Uploaded_Files.
2. In the root folder (DeepFake_Detection), create a folder called model and add the model file in it.

(Paths are already set in server.py, so this avoids errors.)

Results

1. Accuracy of the Model

   

2. Training and Validation Accuracy Graph

   

3. Training and Validation Loss Graph

   

4. Confusion Matrix

   

Developer

• Ajad