A full-stack project for No-Reference Image Quality Assessment (NR-IQA).
Built with Python, PyTorch, FastAPI & Docker, this pipeline takes raw images, runs a trained deep-learning model, and predicts a quality score without needing the original reference image.
Real-world images often suffer from compression, blur, noise, or artifacts.
Traditional IQA metrics (PSNR, SSIM) need a “perfect” reference image — which we rarely have.
This project implements deep learning NR-IQA to:
- Predict perceptual quality from any image
- Enable automated QA in photo platforms, streaming, and dataset filtering
Dataset used: KonIQ-10k (10,000 diverse images, 512×384 px)
- 🚀 FastAPI REST endpoint for real-time scoring
- 🧠 PyTorch model with configurable backbone
- 🎨 Modern UI for uploading images & visualizing scores
- 📦 Docker-ready for clean deployment
- ☁️ Render-friendly: automatic builds & deploys
- 📊 Hooks for model retraining & logging
The /ui endpoint serves a clean, user-friendly web interface for image uploads & quality scoring.
Image_Quality_Assessment/
│
├── artifacts/ # Saved models (model.pth etc.)
├── data/ # (ignored by git) dataset / images
├── src/
│ ├── model.py # PyTorch model & loading utilities
│ ├── inference_api.py # FastAPI app exposing /predict + UI
│ └── utils.py # Preprocessing & helpers
│
├── static/ # HTML, CSS, JS for UI
│ └── index.html
│
├── requirements.txt # Python dependencies
├── Dockerfile # Build instructions for Docker
├── .dockerignore # Ignore unnecessary files for container
├── .gitignore # Ignore venv, data, logs etc.
├── README.md # You’re reading it
└── LICENSE
```bash git clone https://github.com/DarainHyder/Image_Quality_Assessment.git```
```bash cd Image_Quality_Assessment ```
```bash python -m venv venv ```
```bash source venv/bin/activate # (Windows: venv\Scripts\activate) ```
```bash pip install -r requirements.txt ```
🚀 Usage Run API locally
```bash uvicorn src.inference_api:app --host 0.0.0.0 --port 8000 --reload ```
API Docs → http://127.0.0.1:8000/docs
UI Page → http://127.0.0.1:8000/ui
docker build -t image-quality-assessment .
```bash docker run -d -p 8000:8000 image-quality-assessment ```
Visit:
API Docs → http://localhost:8000/docs
UI Page → http://localhost:8000/ui
Python 3.11
PyTorch – Deep learning framework
FastAPI – High-performance web API
Uvicorn – ASGI server
Docker – Containerization & cloud deployment
HTML/CSS – Custom UI frontend
MIT – feel free to fork & remix. If you use this repo, star ⭐ it and tag me — love seeing real-world uses!
Clean code, reproducible pipelines, and a Docker-first mindset — ship models like products. 🚀