AI Photo Studio

AI Photo Studio is a scalable, production-ready image processing service powered by deep learning models. It provides a complete end-to-end solution featuring user authentication, asynchronous task processing, real-time progress monitoring, and cloud-ready storage integration. Currently supports AI-powered background removal using U2Net, with an extensible architecture designed to accommodate multiple AI models and processing types.

Demo

Demo link: http://3.16.188.226/

Architecture Overview

                                   ┌─────────────────┐
                                   │   PostgreSQL    │
                                   │    Database     │
                                   └────────▲────────┘
                                            │ R/W
                                   ┌────────┴────────┐
                                   │                 │
                         ┌─────────┴────────┐        │
                         │                  │        │
┌─────────────┐          │  FastAPI Web     │        │
│   Client    │◀────────▶│    Service       │        │
│  (Browser)  │   HTTP   │                  │        │
└──────▲──────┘          └─┬────────┬───────┘        │
       │                   │        │                │
       │ SSE               │        │                │
       │(via Web)          │Enqueue │Subscribe       │
       │                   │        │                │
       └───────────────────┼────────┼────────────────┘
                           │        │
                           ▼        ▼
                    ┌─────────────┐          ┌─────────────┐
                    │    Redis    │          │   Storage   │
                    │   Queue +   │          │  (AWS S3)   │
                    │   Pub/Sub   │          └──────▲──────┘
                    └──────▲──────┘                 │
                           │                        │ R/W
                           │Dequeue                 │
                           │& Publish               │
                    ┌──────┴────────────────────────┴──┐
                    │         Worker Process           │
                    │  ┌────────────────────────────┐  │
                    │  │   Model Orchestrator       │  │
                    │  │  ┌───────────────────────┐ │  │
                    │  │  │ AI Models (U2Net...)  │ │  │
                    │  │  └───────────────────────┘ │  │
                    │  └────────────────────────────┘  │
                    └──────────────┬───────────────────┘
                                   │ Update Status
                                   ▼
                              PostgreSQL

Key Features / Highlights

Distributed Architecture

• Task Orchestration Framework: Model orchestrator pattern that manages lifecycle of multiple AI models and task types concurrently
• Asynchronous Task Queue System: Redis-based task queue with configurable worker concurrency, enabling horizontal scaling of processing workers
• Event-Driven Real-Time Updates: Server-Sent Events (SSE) with Redis Pub/Sub for live task progress streaming to clients
• Multi-Storage Backend Support: Abstract storage layer supporting both local filesystem and AWS S3, easily extensible to other cloud providers

Reliability & Performance

• Automatic Retry Mechanism: Failed tasks are automatically re-queued with configurable retry policies
• Rate Limiting & Security: Multi-tier API rate limiting (strict/moderate/lenient) to prevent abuse and ensure service stability

Production-Ready Features

• User Authentication System: Complete user registration, login, and session management with Argon2 password hashing
• Task Lifecycle Management: Full task state tracking (PENDING → PROCESSING → COMPLETED/FAILED) with database persistence
• File Validation & Security: File type and size validation, secure file handling with UUID-based naming
• Responsive Web Interface: Modern HTML/CSS/JavaScript frontend for file uploads and task monitoring

Technical Stack

Backend: FastAPI, SQLAlchemy, PostgreSQL, Redis
AI/ML: U2Net (Background Removal), PyTorch
Storage: AWS S3 (aioboto3)
Deployment: Docker Compose
Security: Argon2, JWT/Session-based authentication, API rate limiting

How to Run

Set up environment variable and aws.

Start Service by:

uvicorn app.main:app --host 0.0.0.0 --port YOUR_PORT

and start one worker by:

python -m worker.main

NEXT (Further Development Plan)

Model & Features

• Add more AI models:
  • Style transfer
  • Object detection and segmentation
  • Face detection and beautification
• AWS Lambda integration for serverless worker execution
• Add support for custom model parameters per request
• Add webhook support for task completion notifications

Developer & Operations

• Create admin dashboard for system monitoring
• Implement usage analytics and billing system

Infrastructure & Scalability

• Add Prometheus metrics and Grafana dashboards for monitoring
• Amazon SQS as alternative queue backend for better cloud integration
• Kubernetes deployment configuration

---

Tech Stack: Python • FastAPI • PostgreSQL • Redis • Docker • PyTorch • AWS S3

License: MIT