gitmyhub

turicreate

C++ ★ 11k updated 2y ago ▣ archived

Turi Create simplifies the development of custom machine learning models.

Apple's Python toolkit for building machine learning models without needing to know the math, trains image classifiers, sound recognizers, and recommenders in a few lines of code and exports them as Core ML files ready for iOS or macOS apps.

PythonC++Core MLpipsetup: moderatecomplexity 3/5

Turi Create is a Python toolkit from Apple that lets you build machine learning models without needing to understand the underlying math or algorithms. You work with tasks, not theory: tell it you want to classify images, and it handles the rest. The output is a model file in Apple's Core ML format, which you can drop directly into an iOS, macOS, watchOS, or tvOS app.

The toolkit covers a wide range of common tasks. You can train a model to recognize objects in photos, classify sounds, detect activities from sensor data, group similar images, analyze the sentiment of text, or build a recommendation system that personalizes choices for users. Each task has its own simple function, and the README shows that a working image classifier can be written in about five lines of Python.

Turi Create runs on macOS, Linux, and Windows (via the Windows Subsystem for Linux). It does not require a graphics card, though certain tasks run significantly faster when one is available. Apple GPU acceleration on macOS is automatic. On Linux, enabling GPU support requires following a separate guide included in the repository.

Installation is done through Python's standard package manager with a single pip command. Apple recommends using a virtual environment to keep the package isolated from your other Python projects. The documentation includes a full user guide and API reference hosted on Apple's developer site.

Where it fits