Azure AI Services

Python Vision SDK

There is a notebook exploring and formatting the response

Structure

.
├── notebooks
│   ├── imgs
│   └── Vision-SDK-READ_API.ipynb
└── src
    ├── app_gr.py
    ├── main.py
    ├── output_images
    ├── __pycache__
    │   ├── utils.cpython-310.pyc
    │   ├── utils.cpython-311.pyc
    │   ├── validate.cpython-311.pyc
    │   ├── vision_functions.cpython-310.pyc
    │   └── vision_functions.cpython-311.pyc
    ├── sample.jpg
    ├── temp_images
    ├── utils.py
    ├── validate.py
    └── vision_functions.py
8 directories, 20 files

vision_functions.py has a VisionFunctions Class with diferent azure vision api functions, like OCR, Face detection etc.

Once the resources are provisioned

• Create a venv
• install dependencies
• rename .env.example to .env and add the key and endpoint of the vision resource
• run main.py or app_gr.py for UI

Gradio UI

Provision Azure Resources

First we create the AI Vision resource on our Azure Subscription

We log in

az login

Let's define some variables to not type as much

rg="ai-vision-rg"
vision="ai-vision-01"
loc="West US"

Now we create the resource group

az group create \
	--name $rg \
	--location $loc

We then create the resource

az cognitiveservices account create \
	--name $vision \
	--resource-group $rg \
	--kind ComputerVision \
	--sku F0 \
	--location $loc

Now we need to retrieve the key and endpoint to use later

az cognitive services account keys list \
	--name $vision \
	-g $rg \
	&& \
	az cognitive services account show \
	--name $vision \
	-g $rg \
	--query "properties.endpoint"

Set the environment variables for the key and endpoint

export VISION_KEY="<key>"
export VISION_ENDPOINT="<endpoint>"

Alternative, we can just put all together on a simple script and run that

#!/bin/bash

# Variables
rg="ai-vision-rg"
vision="ai-vision-01"
loc="West US"

# Create resource group
az group create --name $rg --location $location

# Create Vision resource
az cognitiveservices account create \
    --name $vision \
    -g $rg \
    --kind ComputerVision \
    --sku S0 \
    --location $location

# Retrieve and set key, format output 
VISION_KEY=$(az cognitiveservices account keys list \
 --name $vision \
 --resource-group $rg \
 --query "key1" -o tsv)
export VISION_KEY

# Retrieve and set endpoint, also format output
VISION_ENDPOINT=$(az cognitiveservices account show \
 --name $vision \
 --resource-group $rg \
 --query "properties.endpoint" -o tsv)
export VISION_ENDPOINT

# Confirm values
echo "VISION_KEY: $VISION_KEY"
echo "VISION_ENDPOINT: $VISION_ENDPOINT"

We make the script executable and run it

chmod +x create-vision-resources.sh
./create-vision-resources.sh

If we don't need the resource anymore we destroy it

az group delete \ 
--name $rg \
--yes

Verify

az group show --name $rg

Caution

It is not safe to echo secrets on the terminal since there is a history