MobileNetV2 for fastai V1 (pretrained and with [optional] batchnorm fusion)
Usage (see the Jupyter Notebook for details):
from mbnetv2 import mbnetv2, _mbnetv2_split
learn = cnn_learner(data, mbnetv2, cut=-1, split_on=_mbnetv2_split, metrics=error_rate)
learn.fit_one_cycle(4)
Sources: