Workshop: Accelerating Apache MXNet Models on Apple Platforms Using Core ML

Introduction

This repository contains the Python code, deep learning models and the associated jupyter notebook to be used by the participants of the deep learning workshop at re:Invent 2017 (MCL311).

With the release of Core ML by Apple at WWDC 2017, iOS,m acOS, watchOS and tvOS developers can now easily integrate a machine learning model into their app. This enables developers to bring intelligent new features to users with just a few lines of code. Core ML makes machine learning more accessible to mobile developers. It also enables rapid prototyping and the use of different sensors (like the camera, GPS, etc.) to create more powerful apps than ever.

Members of the MXNet community, including contributors from Apple and Amazon Web Services (AWS), have collaborated to produce a tool that converts machine learning models built using MXNet to the Core ML format. This tool makes it easy for developers to build apps powered by machine learning for Apple devices. With this conversion tool, you now have a fast pipeline for your deep-learning-enabled applications. You can move from scalable and efficient distributed model training in the AWS Cloud using MXNet to fast run-time inference on Apple devices.

In this workshop, we will use models trained on millions of examples for one dataset, and apply them to improve performance on a new problem with a much smaller dataset to infer whether an image is that of a cat or a dog. We will then convert this model from MXNet into CoreML and import it into a sample iOS app written in Swift. The iOS app feeds a picture to the model, which predicts whether the image we are looking at is a cat or a dog. For performance, we recommend that you run the app on a physical iOS device (e.g., an iPhone) installed with iOS 11, but you can also try it on the simulator that comes with the Xcode 9.0 using the test images provided with the sample iOS app.

Instructions

1. Download instructions and artifacts:

• Download the git project by:

git clone https://github.com/aws-samples/reinvent-fine-tune-mxnet-model-for-ios-coreml.git

2. Create A Training Compute Instance:

In the Amazon EC2 console, launch an instance. For step-by-step instructions, see Launching an AWS Marketplace Instance in the Amazon EC2 User Guide for Linux Instances. As you follow the steps, use the following values:

• Navigate to the EC2 Console in eu-west-1: https://us-west-2.console.aws.amazon.com/ec2/v2/home?region=eu-west-1

• On the Choose an Amazon Machine Image (AMI) page, choose the Community AMIs tab, and search for Deep Learning AMI Ubuntu Linux - 2.4_Oct2017.

• On the Choose an Instance Type page, choose the p2.xlarge instance type.

• On the Configure Instance Details page, do the following: * From the Network drop-down, choose your VPC. * From the Auto-assign Public IP list, choose Enable

• On the Add Storage page, use the default 50GB for storage size.

• On the Add Tags page, add one tag with the key Name and any value (e.g. mcl311).

• On the Configure Security Group page, leave the default settings which should be:

• Protocol (TCP) rule.

  • Type : Custom TCP Rule
  • Protocol: TCP
  • Port Range: 22
  • Source: Anywhere (0.0.0.0/0,::/0)

• Choose an existing or new key pair and click Launch Instances

3. Connect

Once the instance is running (Status Check: 2/2), create an SSH tunnel to your instance. Enter the following into the terminal:

    ssh -L 8888:localhost:8888 -i <YOUR_CERT>.pem ubuntu@<DNS of your EC2 Instance>

4. Starting the Jupyter Notebook

• Start Jupyter Notebook. Enter the following into the terminal:

    jupyter notebook --NotebookApp.iopub_data_rate_limit=10000000000 &

• Copy the URL output from the jupyter notebook startup command and paste it in your browser. The URL looks like http://localhost:8888/SsomeCookieString

• Upload the fine-tuning-catsvsdogs.ipynb notebook file from your git project to the local jupyter context folder.

5. Run, Train & Model

Open notebook fine-tuning-catsvsdogs.ipynb

• Run all cells in the notebook: select Cell menu and then click on Run Cells.

• Go back to the Jupyter directory page, select the checkmark besides the file named coreml.mlmodel and then click Download to download the file on to your machine.

6. Creating the iOS App

We will be using a sample iOS app to test our newly created CoreML model.

• In your local git folder, open the xcodeproj file in XCode.
• Import the coreml.mlmodel file by dragging it into your XCode project folder.
• Clean and build project
• Test your iOS app

---

IMAGE ATTRIBUTION

The following works of great photography were used in the iOS app:

• https://pixabay.com/en/puppy-dog-pet-animal-cute-white-1903313/
• https://pixabay.com/en/puppy-spaniel-dog-animal-young-2940583/
• https://pixabay.com/en/dog-forest-lake-view-dog-human-2939126/
• https://pixabay.com/en/cat-feline-animal-domestic-pet-2938019/
• https://pixabay.com/en/cat-young-animal-curious-wildcat-2083492/
• https://pixabay.com/en/cat-s-eyes-animal-shelter-mieze-2671903/