333i/4xNomosWebPhoto_RealPLKSR-coreml-ios

4xNomosWebPhoto_RealPLKSR for Core ML (iOS)

A Core ML conversion of 4xNomosWebPhoto_RealPLKSR by Philip Hofmann, a 4x RealPLKSR photo restoration model trained on the Nomos-v2 dataset with realistic noise, lens blur, JPEG and WebP recompression. The weights are unchanged. This repository only adds the export.

Files

File	Use
4xNomosWebPhoto_RealPLKSR_512.mlpackage	Source Core ML model program (fp16), loadable with coremltools or Xcode
4xNomosWebPhoto_RealPLKSR_512_coreml_compiled.zip	Pre-compiled .mlmodelc for direct on-device loading with MLModel(contentsOf:)

Model interface

Input	image, RGB, 512 x 512
Output	upscaled, RGB, 2048 x 2048
Scale	4x
Precision	fp16 ML Program
Minimum OS	iOS 16 / macOS 13
Compute units	All (CPU, GPU, Neural Engine)

Larger images are processed as overlapping 512 x 512 tiles that are feathered back together. The model has no minimum size or alignment requirement beyond the fixed 512 tile.

Verification

Core ML output matches the PyTorch reference at 58 dB PSNR or better on the test tiles (maximum per-channel difference of 3/255), which is within fp16 rounding. On an M-series Mac the model runs in about 135 ms per tile with all compute units enabled.

Conversion

Exported with coremltools 9.0 from the original safetensors checkpoint loaded through spandrel. The model is wrapped so that the input is an RGB image scaled to [0, 1] and the output is clamped and scaled back to [0, 255] as an RGB image, which lets Vision return a VNPixelBufferObservation directly.

Attribution and license

Weights by Philip Hofmann, released under CC BY 4.0. Original release: Phips/4xNomosWebPhoto_RealPLKSR and the GitHub release. This conversion carries the same license and credit requirement.