Deepfake Detection Django Application

Requirements

Note: Nvidia GPU is recommended to run the application efficiently.

• CUDA version >= 10.0
• GPU Compute Capability > 3.0

Main requirements:

Python >= 3.8
Django >= 4.0
PyTorch
Torchvision
OpenCV
NumPy

Install all dependencies using:

pip install -r requirements.txt

---

Directory Structure

• ml_app → Contains machine learning and prediction logic in views.py
• project_settings → Django settings and production configuration
• static → CSS, JavaScript, images and frontend assets
• templates → HTML template files
• models → Trained deep learning models
• uploaded_images → Extracted frames from videos
• uploaded_videos → User uploaded videos

Important: Before running the project, create the following folders in the project root:

models
uploaded_images
uploaded_videos

---

Running Application Using Docker

Step 1: Install Docker Desktop and start Docker

Step 2: Run Deepfake Detection Container

docker run --rm --gpus all \
-v static_volume:/home/app/staticfiles/ \
-v media_volume:/app/uploaded_videos/ \
--name=deepfakeapplication deepfake-detection-image

Step 3: Run Nginx Reverse Proxy Container

docker run -p 80:80 \
--volumes-from deepfakeapplication \
-v static_volume:/home/app/staticfiles/ \
-v media_volume:/app/uploaded_videos/ \
deepfake-nginx-proxy

Step 4: Open Application

http://localhost:8000

---

Running Application Locally

Prerequisite

Copy your trained model into the models folder before running the project.

Model naming format:

model_accuracy_frames.pt
Example:
model_87_acc_10_frames.pt

---

Step 1: Create Virtual Environment

python -m venv venv

Step 2: Activate Virtual Environment

venv\Scripts\activate

Step 3: Install Requirements

pip install -r requirements.txt

Step 4: Run Django Server

python manage.py runserver

Open in browser:

http://127.0.0.1:8000/

---

How the System Works

1. User uploads a video
2. Video is converted into frames
3. Faces are extracted from frames
4. Frames are passed to the deep learning model
5. Model predicts whether video is Real or Fake
6. Result is displayed with confidence score

---

Features

• Video Upload
• Frame Extraction
• Face Detection
• Deepfake Prediction
• Confidence Score Display
• Django Web Interface
• GPU Support
• Docker Support

---

Tech Stack

• Python
• Django
• PyTorch
• OpenCV
• HTML, CSS, JavaScript
• Bootstrap
• Docker
• Nginx

---

Author

Deepfake Detection System Django + Deep Learning Project