core-ml-model-conversion

A sample project demonstrating how to convert a PyTorch model to Core ML format using coremltools and integrate it into an iOS app with SwiftUI.

Features

• Model Conversion: Python script to convert MobileNetV2 (PyTorch) to .mlpackage
• Image Classification: SwiftUI app that classifies photos using Vision + Core ML
• Quantization: Float16 and palettization options for model size reduction
• Photo Picker: Select images from the photo library for classification

Components

Path	Description
convert/convert.py	PyTorch → Core ML conversion script with quantization options
convert/requirements.txt	Python dependencies
Sources/CoreMLDemo/ImageClassifier.swift	Core ML + Vision wrapper for image classification
Sources/CoreMLDemo/ContentView.swift	SwiftUI interface with photo picker and results display
Sources/CoreMLDemo/CoreMLDemoApp.swift	App entry point
Tests/CoreMLDemoTests/	Unit tests for ImageClassifier

Requirements

• Python 3.8+
• Xcode 16.0+
• iOS 17.0+

Usage

1. Convert the Model

cd convert
pip install -r requirements.txt
python convert.py

Quantization options:

python convert.py --quantize float16
python convert.py --quantize palettize --nbits 8

2. Run the App

1. Open CoreMLDemo.xcodeproj in Xcode
2. Drag the generated MobileNetV2.mlpackage into the project
3. Select an iOS simulator or device and run

To regenerate the Xcode project from project.yml:

brew install xcodegen
xcodegen generate

License

This project is licensed under the MIT License - see the LICENSE file for details.