#AzureIoT – How to create an Azure IoT module from an Azure #CustomVision project 👀 1/N

Hi !

I’m a big fan of Azure Custom Vision, and when I talk about Computer Vision I usually share an example on how to tag images from drawings. I had 3 types of drawings

• fish
• flower
• human

I build a model in Custom Vision and then I show my amazing drawing skills with something like this. The custom CV model identifies this as a fish, so I get to a point where my drawing skills are a good fit for this model.

Note: I think the 1st time we did something with these drawings was in an Insider Dev Tour on 2017 or 2018. This photo is from Microsoft Ignite 2019.

Azure Custom Vision is a super cool tool that allow us to create complex Computer Vision projects without code. And, on top of that we have an additional feature:

Once we train a CV model, we can export the model to be used in different platforms. In example, TensorFlow, ONNX, or as a packaged website in a Docker image.

If we go to the Docker option, we will have also the choice to export to different versions:

• Linux
• Windows
• ARM (Raspberry Pi3)

That’s cool. I can download and install the Linux version, and then build and run the project with these 2 commands

docker build -t drawingsit01 .
docker run -p 127.0.0.1:80:80 -d drawingsit01

Now I can test this with a tool like Postman. In order to do this, I draw (again) an amazing fish, and then I perform a POST request to my local running docker container, and it was recognized as a fish with a 99% of accuracy.

So, let’s recap for future posts.

• We created and trained a project in Azure Custom Vision.
• We exported the project as a docker image.
• We run the image locally, and tested with an image.

Right now, I think I have everything that I need to create an Azure IoT Module based on the exported Linux module from Custom Vision.

More in next posts !

Resources

• Azure Custom Vision
• Azure Custom Vision Documentation

¿Con ganas de ponerte al día?

En Lemoncode te ofrecemos formación online impartida por profesionales que se baten el cobre en consultoría:

• Si tienes ganas de ponerte al día con Front End (ES6, Typescript, React, Angular, Vuejs…) te recomendamos nuestros Máster Front End: https://lemoncode.net/master-frontend#inicio-banner
• Si te quieres poner al día en Backend (stacks .net y nodejs), te aconsejamos nuestro Bootcamp Backend: https://lemoncode.net/bootcamp-backend#bootcamp-backend/banner
• Y si tienes ganas de meterte con Docker, Kubernetes, CI/CD…, tenemos nuestro Bootcamp Devops: https://lemoncode.net/bootcamp-devops#bootcamp-devops/inicio