---
license: other
license_name: non-commercial-research
license_link: https://github.com/WuZongWei6/Pixelization
library_name: coreml
pipeline_tag: image-to-image
tags:
  - coreml
  - core-ml
  - ios
  - macos
  - apple
  - on-device
  - pixel-art
  - style-transfer
  - siggraph-asia-2022
---

# Pixelization — Core ML

*Cell-Controllable Pixel Art, SIGGRAPH Asia 2022*

Turn any photo into pixel art. Aliasing-aware generator + anti-alias refinement. Drag the cell-size slider (2–8) to change pixel block size — the network runs once per photo, the slider only re-snaps the grid. 512×512 input. Non-commercial research use only.

Core ML conversion of [WuZongWei6/Pixelization](https://github.com/WuZongWei6/Pixelization) for on-device inference on iPhone, iPad and Mac. Converted with `coremltools`; the packages are stateless, so all sequencing and buffering lives in your Swift code.

| | |
|---|---|
| Task | image to image |
| Upstream | [WuZongWei6/Pixelization](https://github.com/WuZongWei6/Pixelization) |
| Packages | 1 |
| Download size | 35 MB |
| Minimum iOS | 17.0 |
| Peak RAM | ~250 MB |

## Files

| File | Size | Compute units | SHA-256 |
|---|---:|---|---|
| `Pixelization_512.mlpackage.zip` | 35 MB | `cpuAndNeuralEngine` | `f9eac7e8fa6487a4…` |
| **Total** | **35 MB** | | |

`compute_units` is not a suggestion -- it is the configuration the conversion was verified against. Moving a package to a different compute unit can silently change the numerics (FP16 attention overflow) or crash on the GPU.

## Download

```bash
hf download mlboydaisuke/coreml-zoo --include "pixelization/*" --local-dir ./pixelization
unzip './pixelization/pixelization/*.zip' -d ./pixelization
```

## Use in Swift

```swift
import CoreML

let config = MLModelConfiguration()
config.computeUnits = .cpuAndNeuralEngine   // as converted — see the table above

// Unzip the .mlpackage, drop it into your Xcode target and Xcode compiles it
// at build time:
let model = try Pixelization_512(configuration: config)

// ...or compile a downloaded .mlpackage at runtime:
let compiled = try await MLModel.compileModel(at: mlpackageURL)
let model = try MLModel(contentsOf: compiled, configuration: config)
```

## Demo

- **Sample app** — [`sample_apps/PixelizationDemo`](https://github.com/john-rocky/CoreML-Models/tree/master/sample_apps/PixelizationDemo), a standalone SwiftUI project.
- **Models Zoo** — this model is downloadable and runnable inside the [Models Zoo app](https://apps.apple.com/app/id6762083207) on the App Store, no build required.

## Conversion

- Script: [`convert_pixelization.py`](https://github.com/john-rocky/CoreML-Models/blob/master/conversion_scripts/convert_pixelization.py)
- Pitfalls hit during conversion (FP16 overflow, ANE buffer limits, stride handling): [`docs/coreml_conversion_notes.md`](https://github.com/john-rocky/CoreML-Models/blob/master/docs/coreml_conversion_notes.md)
- Model index: [CoreML-Models](https://github.com/john-rocky/CoreML-Models)

## License

The conversion inherits the upstream license: **Research use only**.
See [https://github.com/WuZongWei6/Pixelization](https://github.com/WuZongWei6/Pixelization).

> Upstream ships no LICENSE file; the paper repo states research use only.

## Credits

- Upstream authors: [WuZongWei6/Pixelization](https://github.com/WuZongWei6/Pixelization), 2022
- Core ML conversion: john-rocky (Daisuke Majima)