Tflite model maker custom model transfer

General Discussion
,
#1

I would like to make transfer learning with my own traine custom model:

from tflite_model_maker import model_spec
from tflite_model_maker import object_detector
import tensorflow as tf

train_data = object_detector.DataLoader.from_pascal_voc(

"path to images,
"path to xml,
['label', 'label2']

)

val_data = object_detector.DataLoader.from_pascal_voc(
"path to images,
"path to xml,
[‘label’, ‘label2’]
)

spec = model_spec.get(‘efficientdet_lite1’)

model = object_detector.create(train_data, model_spec=spec, batch_size=8, train_whole_model=True, epochs=1000, validation_data=val_data)

model.export(export_dir=‘.’, tflite_filename=‘model.tflite’)

What I have tried is to change the spec line with my own model as below:

spec = model_spec.get(‘my_model_path’)

But I have an error and it is not working. Does anybody have an idea if I was wrong with this way? If so, What way should I have needed for the aim?

#3

@Chris_Toward,

Welcome to the Tensorflow Forum!

It is not possible to use custom model using tflite_model_maker. Please refer to the supported object detection models in below

Thank you!