This is a multi-GPUs Tensorflow implementation of MobileNetV3 architecture as described in the paper Searching for MobileNetV3.
paper V2 changed some layers from paper V1.
The implementation of paper V1 see branch paper_v1 in this repository for detail.
Tested on tf1.3.0, tf1.10.0, python3.5.
from mobilenet_v3 import mobilenet_v3_large, mobilenetv3_small
model, end_points = mobilenet_v3_large(input, num_classes, multiplier=1.0, is_training=True, reuse=None)
model, end_points = mobilenet_v3_small(input, num_classes, multiplier=1.0, is_training=True, reuse=None)