Deploy production grade compter vision applications using Azure Functions
Needs a location
Details
I had prepared this for the Ignite session, but since we are not conducting that session, I thought to present this topic at our meetup. This was also the topic of our APAC Community session today.
In this session, we will learn more about the Custom Vision service and Azure Functions. We will then train a state of the art custom vision model. We will deploy the model using Azure Functions. In this session, you will learn about the tips and tricks to convert the custom vision model into something that you can deploy.
700pm - 730pm: Demo
730pm - 830pm: Hands on session
This will be an online only meetup. We will use the teams meeting for this session:
https://teams.microsoft.com/l/meetup-join/19%3ameeting_ODdhMjhkNTctM2I5My00Yjg2LTlmNzMtYjViZmE4MWJjNGFm%40thread.v2/0?context=%7b%22Tid%22%3a%2288eabae2-c951-496d-b2a8-aa9bf4b0db92%22%2c%22Oid%22%3a%220fc1c71a-8e31-4a6b-abac-13cf0829062e%22%2c%22IsBroadcastMeeting%22%3atrue%7d
Requirement for Hands-on
1. Azure Account: Make sure you have a working Azure account. Azure functions will remain free forever, while Custom Vision (https://www.customvision.ai/) is free for the first 12 months. I dont have coupons that I can share at this time.
2. Azure Function Core Tools: The installation guide is here: https://docs.microsoft.com/en-us/azure/azure-functions/functions-run-local#v2 I have used chocolatey to install, the steps can be found in the readme https://github.com/Azure/azure-functions-core-tools/blob/master/README.md#windows
3. Visual Studio Code (https://code.visualstudio.com/) with Azure Functions Extensions (https://docs.microsoft.com/en-us/azure/azure-functions/functions-develop-vs-code?tabs=nodejs#install-the-azure-functions-extension)
4. Python 3.X: Prefer to have anaconda installed. You can also work with vanilla python (https://www.python.org/downloads/release/python-374/)
5. Python Packages:
a. Tensorflow 2.0 https://www.tensorflow.org/install
b. numpy==1.17.3
c. pillow
d. flask
6. Suggested Dataset:
a. Fruits 360: https://www.kaggle.com/moltean/fruits
b. Birds 120: https://www.kaggle.com/gpiosenka/100-bird-species
Both the suggested datasets are large files, if you are running short on the space, you can keep just the two classes (folders) and delete the rest of the files. I have created the sample dataset with the Birds 120 containing the two classes of eagles. You can download it from here https://1drv.ms/u/s!AmjwdE_MMESkgZZGe-bOTBPShNdcjQ?e=bEA9t9
