Ever wondered what breed that dog or cat is? In this show, you’ll learn how to train, optimize and deploy a deep learning model using Azure Notebooks, Azure Machine Learning Service, and Visual Studio Code using Python. Using transfer learning to retrain a mobilenet model via Tensorflow to recognize dog and cat breeds using the Oxford IIIT Pet Dataset.

Next, watch how to optimize that model using the Azure Machine Learning Service HyperDrive service, and improve the accuracy of our model to over 90%. Finally, we’ll put on our developer hat, and use Visual Studio Code and our Python Extension to deploy and test our model. Along the way you’ll see cool features like our new Jupyter-powered interactive programming experience in VS Code, our AI powered IntelliSense feature called Intellicode, and our Azure Machine Learning extension.

HyperDrive service, and improve the accuracy of our model to over 90%. Finally, we’ll put on our developer hat, and use Visual Studio Code and our Python Extension to deploy and test our model. Along the way you’ll see cool features like our new Jupyter-powered interactive programming experience in VS Code, our AI powered IntelliSense feature called Intellicode, and our Azure Machine Learning extension.

Github repo for all code used in the show: https://github.com/microsoft/connect-petdetector

Blog post introducing the new features in Azure Notebooks: https://github.com/Microsoft/AzureNotebooks/wiki/Azure-Notebooks-at-Microsoft-Connect()-2018

Blog post introducing our data science features in our Python extension: https://blogs.msdn.microsoft.com/pythonengineering/2018/11/08/data-science-with-python-in-visual-studio-code/

Azure Notebooks: https://notebooks.azure.com

Python Extension: https://marketplace.visualstudio.com/items?itemName=ms-python.python

Azure Machine Learning Extension: https://marketplace.visualstudio.com/items?itemName=ms-toolsai.vscode-ai

Visual Studio Code: https://code.visualstudio.com/

Here’s an interesting talk from PyCon Germany by Joshua Görner, a Data Scientist at BMW.

From the video description:

Interactive notebooks like Jupyter have become more and more popular in the recent past and build the core of many data scientist’s workplace. Being accessed via web browser they allow scientists to easily structure their work by combining code and documentation. Yet notebooks often lead to isolated and disposable analysis artifacts. Keeping the computation inside those notebooks does not allow for convenient concurrent model training, model exposure or scheduled model retraining. Those issues can be addressed by taking advantage of recent developments in the discipline of software engineering. Over the past years containerization became the technology of choice for crafting and deploying applications. Building a data science platform that allows for easy access (via notebooks), flexibility and reproducibility (via containerization) combines the best of both worlds and addresses Data Scientist’s hidden needs.

Jupyter notebooks are great. They are interactive, customizable and can be made to beautifully illustrate data.

Unfortunately only a small fraction of data scientists takes the full advantage of the possibilities that they bring. In this talk, Jakub Czakon shows you some of the coolest notebook features that will impress your peers, dazzle your clients and make your work a lot more enjoyable.