Image Upscaling API

A high-performance FastAPI application for image enhancement and upscaling using RealESRGAN models.

Overview

This project provides a REST API for image upscaling and enhancement using state-of-the-art deep learning models. It supports multiple scaling factors (1x, 2x, 4x, and 8x) and handles various input image formats.

Features

• Multiple Upscaling Options: Scale images by 1x (clarity enhancement), 2x, 4x, and 8x
• Format Conversion: Automatically converts various image formats to optimized JPG
• High-Quality Output: Produces high-quality enhanced images with preserved details
• GPU Acceleration: Utilizes CUDA when available for faster processing
• RESTful API: Simple HTTP interface for easy integration

Technology Stack

• FastAPI: High-performance web framework
• PyTorch: Deep learning framework
• RealESRGAN: State-of-the-art image upscaling models
• OpenCV: Image processing operations
• PIL/Pillow: Image manipulation

Installation

Prerequisites

• Python 3.7+
• CUDA-compatible GPU (optional, for faster processing)

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/image-upscaling-api.git
   cd image-upscaling-api

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Download model weights:

   mkdir -p weights
   # Download RealESRGAN weights to the weights directory
   # RealESRGAN_x2.pth, RealESRGAN_x4.pth, RealESRGAN_x8.pth

Usage

Starting the Server

cd backend
python main.py

The API will be available at http://localhost:8000.

API Endpoints

Endpoint	Method	Description
/rescaling_1x	POST	Enhance image clarity without upscaling
/rescaling_2x	POST	Upscale image by 2x
/rescaling_4	POST	Upscale image by 4x
/rescaling_8x	POST	Upscale image by 8x

Example Request

Using cURL:

curl -X POST -F "image=@./inputs/sample_low_res.jpg" http://localhost:8000/rescaling_4 --output enhanced_image.jpg

Using Python requests:

import requests

url = "http://localhost:8000/rescaling_4"
files = {"image": open("./inputs/sample_low_res.jpg", "rb")}
response = requests.post(url, files=files)

with open("enhanced_image.jpg", "wb") as f:
    f.write(response.content)

Sample Results

Input	2x Upscaling	4x Upscaling

Technical Details

Image Processing Pipeline

1. Input Validation: Verify the uploaded file is a supported image format
2. Format Conversion: Convert the input to a standardized JPG format
3. Model Selection: Choose the appropriate RealESRGAN model based on the requested scale
4. Inference: Process the image through the neural network
5. Response: Return the enhanced image as a JPEG

Models

The application uses RealESRGAN models for super-resolution:

• 1x: Custom OpenCV-based sharpening filter
• 2x: RealESRGAN_x2.pth
• 4x: RealESRGAN_x4.pth
• 8x: RealESRGAN_x8.pth

Project Structure

.
├── backend/
│   ├── main.py                # FastAPI application and endpoints
│   ├── convert_to_jpg.py      # Image format conversion utilities
│   └── RealESRGAN/            # RealESRGAN model implementation
├── weights/                   # Model weights directory
│   ├── RealESRGAN_x2.pth
│   ├── RealESRGAN_x4.pth
│   └── RealESRGAN_x8.pth
├── inputs/                    # Sample input images
├── results/                   # Sample result images
└── README.md                  # This documentation

Known Issues

• The 8x upscaling model is loaded with scale=2 in the code, which may need correction
• Processing large images may require significant memory

License

MIT License

Acknowledgements

• RealESRGAN for the super-resolution models
• FastAPI for the web framework