CoreMLHelpers is a well-known open-source Swift library that makes working with Apple’s Core ML framework noticeably easier, collecting the small types and functions every on-device ML app ends up needing. Maintained by longtime iOS ML author Matthijs Hollemans, it has about 1,450 stars on GitHub.
Running a Core ML model in an app involves a lot of glue code: converting camera images into CVPixelBuffer inputs, turning MLMultiArray outputs back into images or predictions, and post-processing detections. CoreMLHelpers packages those chores into reusable, documented helpers. It converts images to CVPixelBuffer objects and back, converts MLMultiArray data into images, and provides handy functions for common output handling such as top-5 predictions and argmax. For object detection models it includes non-maximum suppression for cleaning up bounding boxes. There is also an experimental, more Swift-friendly wrapper around MLMultiArray for developers who find the raw API awkward. Each feature is documented in its own guide within the repository, making the library as much a learning resource as a dependency. The code was last updated in 2022 but remains a standard reference that iOS developers still reach for when integrating Core ML models.
CoreMLHelpers is free and open source under the MIT license. There is nothing to buy or subscribe to; you add the Swift code to your project and use it under the license terms.
iOS and macOS developers shipping apps with on-device machine learning through Core ML, who want proven helper code for the input/output plumbing instead of rewriting pixel buffer and multi-array conversions for every model.