2xParimgCompact

Tiny, very fast 2× photo upscaler.

A compact 2× photo model trained on Microsoft's ImagePairs dataset. At 2.4 MB it upscales a tile in about a quarter of a second on CPU, which makes it the quick option in our apps.

Download ONNX View repository Paper

Super-resolution Computer Vision

Scale: 2×
Parameters: 0.6M
CPU time per 256 px tile: 0.26s
ONNX file: 2.4MB

Example output

Drag the handle to compare. Produced by running this model.

After: Input: a 240 px photo, enlarged with bicubic resizing. Output: the model's 2× result. — Input: a 240 px photo, enlarged with bicubic resizing. Output: the model's 2× result.

Before: Input: a 240 px photo, enlarged with bicubic resizing. Output: the model's 2× result. — Input: a 240 px photo, enlarged with bicubic resizing. Output: the model's 2× result.

Technical details

Architecture: SRVGGNetCompact (Real-ESRGAN compact, 64 features, 16 convs)
Fine-tuned from: 2x-Compact-Pretrain
Training data: Microsoft ImagePairs
Training: 100,000 iterations
ONNX build: FP32, opset 17, any image size
Input: RGB, NCHW, values 0–1
Benchmark: ONNX Runtime CPU, Apple M2 Pro
Original release: 5 May 2023

Good for

Quick previews
Everyday photos
Large batches

Run it in Flutter

This model runs as-is with our open-source flutter_super_resolution package (MIT), which tiles large images and uses Core ML on Apple devices when available. It works on iOS, Android (ARM devices), macOS and Windows.

pubspec.yaml · Dartgeneral2xlow.onnx

dependencies:
  flutter_super_resolution: ^1.1.1

// Load the model, then upscale a ui.Image
final upscaler = FlutterUpscaler(tileSize: 128, overlap: 8);
await upscaler.initializeModel('assets/models/general2xlow.onnx');

final result = await upscaler.upscaleImage(
  image,
  scale: 2,
  onProgress: (progress, message) => print('$message ${(progress * 100).round()}%'),
);

upscaler.dispose();

Package on GitHub All models that work with it