Checkout Machine Learning features in FileMaker

MonkeybreadSoftware · Saturday at 07:47 AM

Have you seen the machine learning functions in FileMaker 19?

Apple added CoreML framework in 2017 and presented it at WWDC that year. The framework allows to easily load a model for machine learning and call it with your input data. Examples include visual detection of things in pictures like image classification. Other examples include speech detection and sound analysis, text analysis or pattern matching. For example we saw models to take text and tell you the language or the mood of the writer. Also welcome are models to detect porn or violence to filter those.

FileMaker 19 comes built-in with a new script step Configure Machine Learning Model to load a model and let FileMaker know the parameters. Then you can use the new ComputeModel function to run the model. The great thing here is that this is built-in to FileMaker Go, so we expect many more people to use this on the go with FileMaker databases without the need to use iOS SDK and a plugin.

Since 2017 we have CoreML functions in MBS FileMaker Plugin. Compared to the built-in functions in FileMaker, the plugin may have a few more goodies for you. While you can't use it in FileMaker Go, you can do a FileMaker iOS SDK application or in MacOS with FileMaker Pro. Due to the plugin being separate with FileMaker, you can use it in older FileMaker versions if needed. The plugin provides the functionality as functions, so you can use them in custom functions and don't need to use a script. Although we usually use scripts, it can be handy to do something in a Let statement somewhere.

A few extra options you have with our plugin is to provide array parameters to the model. e.g. if you have a model doing math on an one or two dimension array, check the CoreML.AddArrayParameter and CoreML.Add2DArrayParameter functions. Use our CoreML.ResultImage function for models returning you an image. And finally use CoreML.Update to update the model on the go with new data.