HOL01-Cogs

Day#2 – AI & ML

HOL01: AI Apps - COGS-CustomVision-PowerApps

Overview

Create an intelligent application for image classification using AI service. Components:

• Object Clasification model - CustomVision.ai
• PowerApps application to call the object classification serivce and display results

Prerequisitites

• Microsoft Azure account
• Microsoft O365 account (PowerApps access)
• images to classification (at least 15 per category)

Guide

1. CustomVision (https://cusomvision.ai)

1. create CustomVision project - classification (select Multiclass - from Classification types choices)
2. upload images - get data from zip file
3. label / tag images to separate classes within CustomVision portal
4. run training!
5. quick test with one of the test images

You have now created a first model for car classification.

2. Test the created model by callin Custom Vision API

1. Go to https://notebooks.azure.com/ and log in with your Azure credentials.
2. Go to My Projects (upper left)
3. Import Git Repo (azurecz/azuretechacademy-data-analyst-labs-day2)
4. check the import was successfull
5. go to HOL01-Cogs folder and select [Jupyter Notebook] CustomVision-Prediction-API.ipynb in HOL1-Cogs folder and follow the steps.

note: if you see error for missing kernell select Python 3.6

note: change first code cell from !/data/anaconda/envs/py36/bin/pip install azure-cognitiveservices-vision-customvision to !pip install azure-cognitiveservices-vision-customvision

[OPTIONAL] 2. PowerApps (https://powerapps.microsoft.com)

1. create PowerApps application - phone layout

2. add elements: labels, media

3. add CustomVision as datasource [prediction key]

4. select "AddMediaButton1" and put into OnChange action scoring the image (use your CustomVision project ID insted of xxx...): UpdateContext({res:CustomVision.PredictImage("xxxxxxxx-xxxx-xxx-xxxx-xxxxxx", UploadedImage1).Predictions})

5. update label property to update text based on results (text): Concatenate("Probability: ",Text(First(res).Probability))

6. save & publish the app