DeepScan — AI Deepfake Detection

A Flask-based deepfake detection web app that analyzes videos frame-by-frame using a trained EfficientNet-B0 model. It provides real-time progress via SSE and generates a forensic report with charts and flagged evidence.

Prerequisites

• Python 3.9+
• pip

Setup

# 1. Clone the repository
git clone <repo-url>
cd internship

# 2. (Recommended) Create and activate a virtual environment
python -m venv venv
# Windows:
venv\Scripts\activate
# macOS/Linux:
source venv/bin/activate

# 3. Install dependencies
pip install -r requirements.txt

Run the App

python app.py

The server starts at http://localhost:5000. Open it in a browser, upload a video, and the analysis runs automatically.

The pre-trained model deepfake_detector.h5 and DNN face detector files (models/) are already included. No additional setup is required.

Project Structure

├── app.py                  # Flask server (routes, face detection, analysis logic)
├── deepfake_detector.h5    # Trained EfficientNet-B0 model
├── requirements.txt        # Python dependencies
├── models/
│   ├── deploy.prototxt             # DNN face detector config
│   └── res10_300x300_ssd_iter_140000.caffemodel  # DNN face detector weights
├── templates/
│   └── index.html          # Frontend UI
├── static/
│   └── style.css           # Styles
├── uploads/                # Uploaded videos (created at runtime)
├── data/                   # Training/testing datasets
└── Report/                 # Project report

Notes

• The app accepts MP4, AVI, and MOV files up to 500 MB.
• Uploaded videos are analyzed in a background thread; the UI updates in real time.
• The model is loaded at startup — this takes a few seconds.