Add custom (TensorFlow Lite) models support to the ML Kit feature

[EddyVerbruggen]

Checklist for https://firebase.google.com/docs/ml-kit/use-custom-models:

• Use a custom TensorFlow Lite model, on the device
• Use a custom TensorFlow Lite model, stored in Firebase (so users always have the latest model) (having trouble, let's postpone this one)

[EddyVerbruggen]

On iOS I'm experiencing a crash when resizing an image (see the comments in the code). Going to postpone this.

[neil-119]

Awesome! This is going to be great!

[geoffbullen]

Looking forward to this being available!

[OscarAgreda]

Adding Tensor Flow Lite to Eddy's work will be terrific, a game changer for sure.

[efgonz]

This is great but seems that we have to wait a bit, so, is there any other way to load a custom tensorflow model in a nativescript app with angular?

[EddyVerbruggen]

Hi folks, I'm actively working on this task and I'm seeing a few results that give me good hope for the near future ;)

What would be really helpful to me is receiving a few custom tflite models, along with a sample image and expected labels (and scores). Would really appreciate if anyone can provide me with those, otherwise I'll continue development with the single model I have.

[jlooper]

On it!

[jlooper]

Eddy, there are some here https://www.tensorflow.org/lite/models

[geoffbullen]

I’m excited!

[EddyVerbruggen]

I'm wrapping things up for a first release with support for loading and interpreting bundled quantized models on both iOS and Android. Interpretation can be performed from a still image, as well as a live camera feed.

Please check the updated ML Kit docs. I'll release 7.7.0 soon, so y'all can give it a shot! 💪