How to extract Figures with labels from the image

Hi Pavankumar Purilla,

Thank you for Providing Answer, I have used Azure Document Intelligence and got the desired result, but I want to create a bounding box around the figure, as I have also tried out the sample code given on GitHub for extracting bounding box coordinates, but I am unable to track the figure through coordinates as I don't know about the coordinates metrics units
Is there any way to configure bounding box coordinates such that to create bounding box around figure?

Hi Pavankumar Purilla,

Thank you for providing Answer, I have used Azure Document Intelligence, Got the desired result. But I want to create a Bounding box around the figure, for that i have used sample code from GitHub(https://github.com/Azure-Samples/document-intelligence-code-samples/blob/main/Python(v4.0)/Layout_model/sample_analyze_layout.py).

As it is able to provide me coordinates of figure, But I am unable to mark a bounding box around figure because i am unable to know what units/metrics measures were used in coordinates.

Please share some thoughts on how to create bounding box by the resulted coordinates.

Like I have used Document Intelligence Studio, In extracted pages, it is highlighting the figures same as I want to create bounding box

I have given image as reference

Hi Anchit Gupta
Hope you are doing well.
To create a bounding box around a figure the coordinates returned by the service are normalized values in the range [0, 1], relative to the width and height of the image or document analyzed. To draw the bounding box, multiply these normalized coordinates by the actual dimensions of the image (width and height) to convert them into pixel-based absolute coordinates. Once converted, use a library like Pillow or OpenCV to draw the bounding box. For example, if the normalized coordinates are [(x1, y1), (x2, y2), (x3, y3), (x4, y4)], scale each x by the image width and each y by the image height. Then, use these scaled coordinates to draw the box on the image. This approach replicates how Document Intelligence Studio highlights figures.
I hope this information helps. Thank you!

Hi Pavankumar Purilla,
Thanks for your reply,
But I am unable to get the desired result as per your suggestion,

Lets say there is an highlighted figure whos coordinates are:

	"polygon": [
							2.0227,
							4.083,
							6.1379,
							4.0809,
							6.1407,
							7.7027,
							2.0256,
							7.7047
						]

Image dimensions: 1655x2340
If I multiply image dimensions with coordinates of highlighted figure, then resultant coordinates are:
[(3347.5685, 9554.220000000001), (10158.2245, 9549.305999999999), (10162.8585, 18024.318), (3352.368, 18028.998)]

which is out of bound of image size,

Please do explain me how it actually works if I misunderstood.
Thank You!