azure-percept-advanced-development/machine-learning-notebooks

Machine Learning Notebooks

Please note! The experiences in this repository should be considered to be in preview/beta. Significant portions of these experiences are subject to change without warning. No part of this code should be considered stable.

This folder contains some IPython notebooks that show you how to bring a model that you train yourself to the device. Use these instructions as examples to train up a model and bring it to the device so that you don't have to rely on the training that the models have already been through (which will usually be too general for your use case).

Note: These notebooks use models that have already been implemented in the azureeyemodule. If you want to bring a completely custom model to the device, you will need to add support for your custom model into the azureeyemodule and deploy the custom azureeyemodule to your device. If you port a well-known model that others will likely want to use, please consider opening a pull request! See the azureeyemodule folder for instructions.

Transfer Learning using Single Shot Detector

In this Jupyter notebook, you can see how to do transfer learning to bring a custom SSD network to the device. There are two tutorial options to guide you through working with the notebook:

• Cloud development: in this tutorial, you will run the notebook in the Azure Machine Learning Portal with a remote compute instance.

• Local development: in this tutorial, you will run the notebook locally within VS Code, but using a remote compute instance.

Note: This notebook doesn't use AML features or Docker to manage dependencies. This notebook is a little out-dated. It uses TensorFlow 1.x, rather than 2.x. There is work planned to improve this notebook, but it still works, and for now, please use it as an example of how you might go about doing transfer learning for the device.

Transfer Learning using Single Shot Detector - Integrated with Azure ML

In this folder, you can see how to do transfer learning for an SSD network, as above, but also how to collect custom data from the Percept DK Vision camera and train a custom model on that data with an end-to-end Jupyter notebook experience. You also will learn how to label data in Azure ML Labeling projects.

Note: The model training notebook is a little out-dated. It uses TensorFlow 1.x, rather than 2.x. There is work planned to improve this notebook, but it still works, and for now, please use it as an example of how you might go about doing transfer learning for the device.