Face parsing (BiSeNet) — Core ML

FaceParsing.mlpackage — 19-class facial semantic segmentation, converted to Core ML and re-targeted for macOS.

Classes cover skin, brows, eyes, glasses, ears, earrings, nose, mouth, lips, neck, necklace, clothing, hair and hat — the standard CelebAMask-HQ label set.

Provenance and license

Converted from yakhyo/face-parsing (MIT) — BiSeNet face parsing, PyTorch weights with ONNX export. This repository redistributes it as a Core ML format conversion under the same MIT license. All credit for the model and its training goes to yakhyo.

Use

Runs in any Core ML image pipeline. In AML for TouchDesigner it loads into the CoreML Model operator, which detects the 19-class output and selects its Face Parsing palette automatically.

Note: models compiled for iOS can fail to load on macOS with error -1. This package is already re-targeted for macOS deployment.