Loading tflite_model_maker.object_detector from SAVED_FILE format

TensorFlow General Discussion
, , , ,
1

I’ve been working my way through Retrain EfficientDet-Lite detector for the Edge TPU (TF2) with good success, including training EfficientDet_Lite models with my own data set, evaluating results, etc.

Having created a new model (after several hours of training time), I would like to save it into MyDrive so I can reconstitute it later (in a new colab session), without having recreate it from the training set. I have successfully exported the model in SAVED_FILE format. But I can’t find the magic necessary to restore it from save file bundle.

(note this is a different request than downloading the model.tflite file to a mobile app)
tflite #object_detector model_maker help-request

3

Hi @Robert_Zak, The tflite_model_maker returns a keras model instance hence loading the model will be the same as loading the keras model. For example

tf.keras.models.load_model(path)

Thank You.