Tflite Model Maker Architecture MobileNetv3

TensorFlow General Discussion
,
1

Hi,

I am having training via Tflitemodel maker Python and it uses efficientdet architecture. I would like and need to train a model with MobileNetv3 architecture.

Is there anyone could suggest how to do it?

On standard training it uses below script line:

spec = model_spec.get(‘efficientdet_lite1’ )

but to change it to another architecture seems unpossible. Am I right?

2

@murkoc,

Yes, currently the tflite model maker supports the following architectures

OBJECT_DETECTION_MODELS = [
    'efficientdet_lite0',
    'efficientdet_lite1',
    'efficientdet_lite2',
    'efficientdet_lite3',
    'efficientdet_lite4',
]

IMAGE_CLASSIFICATION_MODELS = [
    'efficientnet_lite0', 'efficientnet_lite1', 'efficientnet_lite2',
    'efficientnet_lite3', 'efficientnet_lite4', 'mobilenet_v2', 'resnet_50'
]

Thank you!

1 Like
3

Thank you for the information.