Awesome

Overview

MobileNet-v2 is a convolutional neural network that has been trained on a subset of the ImageNet database. As a result, the network has learned rich feature representations for a wide range of images. The network can classify images into 1000 object categories, such as keyboard, mouse, pencil, and many animals. The network has been designed to have a low memory footprint, making it ideal for deployment to low-memory hardware.

The network has an image input size of 224-by-224-by-3.

Usage

This repository requires MATLAB (R2019a and above) and the Deep Learning Toolbox.

This repository provides three functions:

mobilenetv2Layers: Creates an untrained network with the network architecture of MobileNet-v2
assembleMobileNetv2: Creates a MobileNet-v2 network with weights trained on ImageNet data
mobilenetv2Example: Demonstrates how to classify an image using a trained MobileNet-v2 network

To construct an untrained MobileNet-v2 network to train from scratch, type the following at the MATLAB command line:

lgraph = mobilenetv2Layers;

The untrained network is returned as a layerGraph object.

To construct a trained MobileNet-v2 network suitable for use in image classification, type the following at the MATLAB command line:

net = assembleMobileNetv2;

The trained network is returned as a DAGNetwork object.

To classify an image with the network:

img = imresize(imread("peppers.png"),[224 224]);
predLabel = classify(net, img);
imshow(img);
title(string(predLabel));

Documentation

For more information about the MobileNet-v2 pre-trained model, see the mobilenetv2 function page in the MATLAB Deep Learning Toolbox documentation.

Architecture

MobileNet-v2 is a residual network. A residual network is a type of DAG network that has residual (or shortcut) connections that bypass the main network layers. Residual connections enable the parameter gradients to propagate more easily from the output layer to the earlier layers of the network, which makes it possible to train deeper networks. This increased network depth can result in higher accuracies on more difficult tasks.

You can explore and edit the network architecture using Deep Network Designer.

MobileNet-v2 in Deep Network Designer

MobileNet-v2 in MATLAB

This repository demonstrates the construction of a residual deep neural network from scratch in MATLAB. You can use the code in this repository as a foundation for building residual networks with different numbers of residual blocks.

You can also create a trained MobileNet-v2 network from inside MATLAB by installing the Deep Learning Toolbox Model for MobileNet-v2 Network support package. Type mobilenetv2 at the command line. If the Deep Learning Toolbox Model for MobileNet-v2 Network support package is not installed, then the function provides a link to the required support package in the Add-On Explorer. To install the support package, click the link, and then click Install.

Alternatively, you can download the MobileNet-v2 pre-trained model from the MathWorks File Exchange, at Deep Learning Toolbox Model for MobileNet-v2 Network.

You can create an untrained MobileNet-v2 network from inside MATLAB by importing a trained MobileNet-v2 network into the Deep Network Designer App and selecting Export > Generate Code. The exported code will generate an untrained network with the network architecture of MobileNet-v2.