Get Area of Interest of image using Azure Cognitive Service

Question

I need to extract the area of interest for images - https://learn.microsoft.com/en-in/azure/cognitive-services/computer-vision/concept-generating-thumbnails#area-of-interest.

I am using python 3.x and wish to upload an image in stream or as a URL. I don't see how to implement this among the guides given on this page - https://learn.microsoft.com/en-in/azure/cognitive-services/computer-vision/quickstarts-sdk/image-analysis-client-library?tabs=visual-studio&pivots=programming-language-python.

computervision_client = ComputerVisionClient(endpoint, CognitiveServicesCredentials(subscription_key))  
remote_image_url = "https://static.hub.91mobiles.com/wp-content/uploads/2020/06/xiaomi-retail-store-et.jpg"  
  
aoi = computervision_client.get_area_of_interest(remote_image_url)  
print(aoi)  

I ran the above code and get the following output:

{'additional_properties': {}, 'area_of_interest': <azure.cognitiveservices.vision.computervision.models._models_py3.BoundingRect object at 0x000001A997A41CA0>, 'request_id': 'dd0d418a-223c-43b7-bc44-e04aafceadb1', 'metadata': <azure.cognitiveservices.vision.computervision.models._models_py3.ImageMetadata object at 0x000001A997A7EA90>, 'model_version': '2021-04-01'}  

How do I go about extracting meaningful data from this? The output mentions - BoundingRect object at 0x000001A997A41CA0 - can this give me the cordinates?

Answer 1

@Vedant Modi The area_of_interest variable contains the co-ordinates the API returned in your image. You can use this co-ordinates to crop the image or create a bounding box using a package like OpenCV.

In your code snippet if you print the details, you should see the co-ordinates for AOI and other metadata of image.

print(aoi.area_of_interest)  
print(aoi.request_id)  
print(aoi.metadata)  

In this case something like this is printed.

{'additional_properties': {}, 'x': 146, 'y': 0, 'w': 450, 'h': 450}  
49499382-63fc-465f-9d46-f21c6148c188  
{'additional_properties': {}, 'width': 600, 'height': 450, 'format': 'Jpeg'}