ML Kit

Think of ML Kit as a set of pre-packaged ML functionalities that you can plug into your app with relatively little code. It sits at a higher level of abstraction compared to directly using a low-level runtime like TensorFlow Lite or ExecuTorch. ML Kit often leverages these runtimes under the hood but provides a simplified API for common mobile ML tasks.

for example, ML Kit provides a pre-trained ready to use object detection model. If this model satisfies the requirements of my problem, great, but if it doesn't I prob have to train my own or see if the ML Kit implementation is customizable enough.

the processing is done on the device and can also leverage the device's GPU or TPU if present.

while ML Kit is a native Android and iOS SDK, there seem to be good third-party Flutter Plugins that wrap the native ML Kit APIs and expose them to your Flutter code.