PyTorch Implementation of MobileNet V3

Reproduction of MobileNet V3 architecture as described in Searching for MobileNetV3 by Andrew Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, Quoc V. Le, Hartwig Adam on ILSVRC2012 benchmark with PyTorch framework.

Requirements

Dataset

Download the ImageNet dataset and move validation images to labeled subfolders. To do this, you can use the following script: https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh

Training recipe

batch size 1024
epoch 150
learning rate 0.4 (ramps up from 0.1 to 0.4 in the first 5 epochs)
LR decay strategy cosine
weight decay 0.00004
dropout rate 0.2 (0.1 for Small-version 0.75)
no weight decay biases and BN
label smoothing 0.1 (only for Large-version)

Models

Architecture	# Parameters	MFLOPs	Top-1 / Top-5 Accuracy (%)
MobileNetV3-Large 1.0	5.483M	216.60	74.280 / 91.928
MobileNetV3-Large 0.75	3.994M	154.57	72.842 / 90.846
MobileNetV3-Small 1.0	2.543M	56.52	67.214 / 87.304
MobileNetV3-Small 0.75	2.042M	43.40	64.876 / 85.498

from mobilenetv3 import mobilenetv3_large, mobilenetv3_small

net_large = mobilenetv3_large()
net_small = mobilenetv3_small()

net_large.load_state_dict(torch.load('pretrained/mobilenetv3-large-1cd25616.pth'))
net_small.load_state_dict(torch.load('pretrained/mobilenetv3-small-55df8e1f.pth'))

Citation

@InProceedings{Howard_2019_ICCV,
author = {Howard, Andrew and Sandler, Mark and Chu, Grace and Chen, Liang-Chieh and Chen, Bo and Tan, Mingxing and Wang, Weijun and Zhu, Yukun and Pang, Ruoming and Vasudevan, Vijay and Le, Quoc V. and Adam, Hartwig},
title = {Searching for MobileNetV3},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}