Azure Form Recognizer & Image Extraction

Question

Azure Form Recognizer & Image Extraction

Karyn Close (Work) 5

Does Azure Form Recognizer have the ability to extract images from a PDF? If not, is there another solution where this can be done?

YutongTie-MSFT 54,016 Reputation points Moderator

2023-06-13T22:59:04.12+00:00

Hello @Karyn Close (Work)

Thanks for reaching out to us, may I know more information about your scenario? Do you mean you want to extract text and text on some images? Could you please share a sample PDF page to me so that we can check?

Regards,

Yutong
Karyn Close (Work) 5 Reputation points

2023-06-13T23:07:58.1933333+00:00

I mean, I have a picture with no text and I want to extract just the image. I'm wanting to extract a pictogram from a safety data sheet and associate it with a chemical.
YutongTie-MSFT 54,016 Reputation points Moderator

2023-06-13T23:18:43.4+00:00
Hello @Karyn Close (Work)

Thanks for your information, if you have a picture with no text and you want to extract just the image, you can use Azure Computer Vision API to perform image analysis and extract the image. Azure Form Recognizer is for text extraction only at this moment.

Reference - https://learn.microsoft.com/en-us/azure/cognitive-services/computer-vision/concept-generating-thumbnails

Except thumbnail, you can also do what you want according to your scenario.

Here are the steps you can follow:

Create a Computer Vision resource in Azure portal and obtain the endpoint and access key.

Use the "Analyze Image" API to analyze the image. Set the "visualFeatures" parameter to "Categories, Tags, Description, Color, ImageType". This will enable the API to detect the image type and provide additional information about the image.

Check the "clipArtType" and "lineDrawingType" properties in the response. If the "clipArtType" is "0" and the "lineDrawingType" is "0", then the image is a photograph and you cannot extract it. If the "clipArtType" is "1" or "2" and the "lineDrawingType" is "0", then the image is a clipart and you can extract it.

Use the "Get Thumbnail" API to extract the image. Set the "width" and "height" parameters to the size of the image you want to extract.

Associate the extracted image with the chemical by storing the image URL or file path in a database or other data store.

Here is an example of what the code might look like in Python:

import requests # Set the API endpoint and access key endpoint = "<your endpoint>" access_key = "<your access key>" # Set the image URL image_url = "<your image URL>" # Set the thumbnail parameters params = { "width": 100, "height": 100, "smartCropping": "true" } # Call the "Generate Thumbnail" API response = requests.get(endpoint + "/generateThumbnail", headers={"Ocp-Apim-Subscription-Key": access_key}, params=params, json={"url": image_url}) response.raise_for_status() # Save the thumbnail to a file with open("thumbnail.jpg", "wb") as f: f.write(response.content)

In this example, the "Generate Thumbnail" API is called with the specified width and height parameters. The thumbnail is then saved to a file named "thumbnail.jpg". You can modify this code to display the thumbnail in your application instead of saving it to a file.

I hope this helps!

Regards,

Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.
Karyn Close (Work) 5 Reputation points

2023-06-15T14:23:50.8033333+00:00

When I walk through your instructions and I get to this part, I don't see the options you are suggesting to fill in.

Use the "Analyze Image" API to analyze the image. Set the "visualFeatures" parameter to "Categories, Tags, Description, Color, ImageType". This will enable the API to detect the image type and provide additional information about the image. Check the "clipArtType" and "lineDrawingType" properties in the response. If the "clipArtType" is "0" and the "lineDrawingType" is "0", then the image is a photograph and you cannot extract it. If the "clipArtType" is "1" or "2" and the "lineDrawingType" is "0", then the image is a clipart and you can extract it.

For Analyze Image connector in Power Automate, I see Image Source, Language, and Image URL. Am I not understanding what you are explaining?

1 answer

Your answer

YutongTie-MSFT 54,016 Reputation points Moderator

2023-06-13T22:59:04.12+00:00

Hello @Karyn Close (Work)

Thanks for reaching out to us, may I know more information about your scenario? Do you mean you want to extract text and text on some images? Could you please share a sample PDF page to me so that we can check?

Regards,

Yutong
Karyn Close (Work) 5 Reputation points

2023-06-13T23:07:58.1933333+00:00

I mean, I have a picture with no text and I want to extract just the image. I'm wanting to extract a pictogram from a safety data sheet and associate it with a chemical.
Karyn Close (Work) 5 Reputation points

2023-06-15T14:23:50.8033333+00:00

When I walk through your instructions and I get to this part, I don't see the options you are suggesting to fill in.

Use the "Analyze Image" API to analyze the image. Set the "visualFeatures" parameter to "Categories, Tags, Description, Color, ImageType". This will enable the API to detect the image type and provide additional information about the image. Check the "clipArtType" and "lineDrawingType" properties in the response. If the "clipArtType" is "0" and the "lineDrawingType" is "0", then the image is a photograph and you cannot extract it. If the "clipArtType" is "1" or "2" and the "lineDrawingType" is "0", then the image is a clipart and you can extract it.

For Analyze Image connector in Power Automate, I see Image Source, Language, and Image URL. Am I not understanding what you are explaining?

Answer 1

Karyn Close (Work) 5

Thank you!

Share via

Azure Form Recognizer & Image Extraction

1 answer

Your answer