4xPurePhoto RealPLSKR

4× photo upscaler that excels at hair, fur and fine detail.

A RealPLKSR model trained on 8,684 hand-picked photos. It is especially good with hair, animals and party shots, and handles inputs as small as about 300 px on a side.

Download ONNX View repository Paper

Super-resolution Computer Vision

Scale: 4×
Parameters: 7.4M
CPU time per 256 px tile: 3.9s
ONNX file: 30MB

Example output

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

After: Input: a 120 px photo of a cat, enlarged with bicubic resizing. Output: the model's 4× result. — Input: a 120 px photo of a cat, enlarged with bicubic resizing. Output: the model's 4× result.

Before: Input: a 120 px photo of a cat, enlarged with bicubic resizing. Output: the model's 4× result. — Input: a 120 px photo of a cat, enlarged with bicubic resizing. Output: the model's 4× result.

Technical details

Architecture: RealPLKSR (partial large kernel CNN)
Training data: 8,684 hand-picked photos
Training: 152,000 iterations
ONNX build: FP32, opset 17, any image size
Input: RGB, NCHW, values 0–1
Benchmark: ONNX Runtime CPU, Apple M2 Pro
Original release: 21 September 2024

Good for

Pets and hair
Portraits
Small photos

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 · Dartart4xhigh.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/art4xhigh.onnx');

final result = await upscaler.upscaleImage(
  image,
  scale: 4,
  onProgress: (progress, message) => print('$message ${(progress * 100).round()}%'),
);

upscaler.dispose();

Package on GitHub All models that work with it