🛡️ Deep Visionary

Vision Transformer • FastAPI • Celery • Redis • React • Docker

A full-stack, production-ready DeepFake image detection system powered by a Vision Transformer (ViT) model. The application supports asynchronous inference, real-time frontend polling, and fully dockerized deployment.

---

🚀 Features

• 🔍 Vision Transformer (ViT) based DeepFake detection
• ⚡ FastAPI backend with async request handling
• 🧵 Celery + Redis for background inference
• 🌐 React (Vite) frontend
• 🐳 Fully Dockerized (Backend, Worker, Redis, Frontend)
• 📡 Async API (/predict → /result/{task_id})
• 🔐 Proper CORS & Docker networking
• 📦 Model persistence using Docker volumes

---

🏗️ Architecture

Frontend (React)
     |
     | HTTP
     v
FastAPI Backend
     |
     | Async Task
     v
Celery Worker
     |
     | Inference
     v
Vision Transformer Model
     |
     v
Redis (Broker & Results)

---

🔌 API Endpoints

➤ Predict Image

POST /predict

Request

• multipart/form-data
• Field: file

Response

{
  "task_id": "abc123"
}

---

➤ Get Prediction Result

GET /result/{task_id}

Processing

{ "status": "processing" }

Success

{
  "status": "done",
  "prediction": {
    "label": "Real",
    "confidence": 0.99
  }
}

Failure

{ "status": "failed" }

---

🐳 Docker Setup

Prerequisites

• Docker
• Docker Compose

---

Build & Run

docker-compose up --build

---

Access Services

Service	URL
Frontend	http://localhost:3000
Backend	http://localhost:8000
Redis	redis://localhost:6379

---

🌍 Docker Networking (Important)

• ❌ localhost does NOT work between containers
• ✅ Containers communicate via service names

Frontend uses:

VITE_API_URL=http://backend:8000

---

🔐 CORS Configuration

Backend allows frontend access:

allow_origins=["http://localhost:3000"]

---

📦 Model Handling

• Model files are NOT committed to GitHub
• Loaded at startup or from Docker volume

volumes:
  - model-data:/app/content/final-vit-model

---

🧠 Why Celery?

• Prevents blocking API requests
• Handles heavy ML inference
• Enables scalable async processing
• Perfect for production ML systems

---

🛠️ Tech Stack

Backend

• FastAPI
• Celery
• Redis
• PyTorch / Transformers

Frontend

• React
• Vite
• Tailwind

DevOps

• Docker
• Docker Compose

---

🚀 Deployment Ready

Can be deployed on:

• AWS (EC2 / ECS)
• Google Cloud (GCE / Cloud Run)
• Azure (VM / Containers)

Next steps:

• Cloud deployment
• CI/CD with GitHub Actions

---

👨‍💻 Author

Muhammad Daniyal Ijaz AI / ML Engineer FastAPI • Vision Transformers • Docker • Cloud

---

⭐ Support

If you find this project useful:

• ⭐ Star the repository
• 🍴 Fork it
• 🚀 Deploy it