Flutter Super Resolution

A Flutter library for upscaling images using ONNX Runtime neural networks with advanced tiling support.

Platform Support

• Windows
• macOS
• iOS
• Android (Only ARM devices)

Features

• High-quality image upscaling using machine learning models
• Supports full image and tiled processing
• Customizable tile size and overlap
• Progress tracking during upscaling
• Optimized for mobile devices

Installation

Add to your pubspec.yaml:

dependencies:
  flutter_super_resolution: ^1.0.0
  onnxruntime: ^latest_version

Usage

final upscaler = FlutterUpscaler(
    tileSize: 128,   // Optional: Customize tile processing
    overlap: 8       // Optional: Prevent tile seams
);

// Initialize model from assets
await upscaler.initializeModel('assets/super_resolution_model.onnx');

// Upscale image
final upscaledImage = await upscaler.upscaleImage(
sourceImage,
scale: 2,        // Scale factor (2x, 4x, etc.)
onProgress: (progress, message) {
print('$message: ${(progress * 100).toStringAsFixed(1)}%');
}
);

Parameters

• tileSize: Size of processing tiles (default: 128)
• overlap: Pixel overlap between tiles to reduce seam artifacts (default: 8)

Methods

• initializeModel(modelPath): Load ONNX model from Flutter assets
• initializeModelFromFile(filePath): Load ONNX model from device file
• upscaleImage(image, scale): Upscale image with optional progress tracking
• dispose(): Release model resources

Requirements

• Flutter SDK
• ONNX Runtime Flutter plugin
• Pre-trained ONNX super-resolution model

Performance Tips

• Use smaller tile sizes for memory-constrained devices
• Adjust overlap to balance processing quality and speed

License

MIT License