tflite model with a Tensor.image as input instead of Tensor.buffer

Question

I\'m new to machine learning and have (maybe a stupid) question. When I create a model for image classification in Keras

train = ImageDataGenerator().flow_fro