I am trying to make an android based app with tflite model.
With tflite_model_maker, the image_classifier does not seem to have any compilation functionality.
I want to add tfa metrics, monitoring metrics, early stopping, etc in my model.
Is there any way to do so or maybe we could use our saved model(H5 ) and load it .
I have tried using tf.lite.TFLiteConverter.from_keras_model(model).But there at a point I faced issues of tensorBuffer in tflite model.
The output of Model Maker can be a regular saved_model file and with that you can load and calculate additional metrics but I don’t think you can use early stoping
To use the TFLite model, I’d suggest you look into Task Library, it makes everything MUCH easier: TensorFlow Lite Task Library