We are using
Tensorflow 1.14
Keras 2.1.2
GPU: GeForce GTX 1660 Ti major: 7 minor: 5 memoryClockRate(GHz): 1.86
for custom object detection using Mask-RCNN from this repo https://github.com/matterport/Mask_RCNN.
Now we trained a model successfully and its detecting objects on our desktop. Now, we want to generate tflite for mobile usage where we are facing below mentioned error:
ValueError: Unknown layer BatchNorm
Please note that we have created weights and keras model .h5 with different scripts
We have tried following code to convert keras model to tflite
import tensorflow as tf
converter = tf.lite.TFLiteConverter.from_keras_model_file( 'Save-Model8.h5')
tfmodel = converter.convert()
open ("model.tflite", "wb") .write(tfmodel)