0

Can we send specific coordinates within an input image to the MS handwriting recognition API so that not the whole but only those specific areas in the image are recognized.

Abhinav Mohan
  • 1
  • 1
  • Are you able to crop the image client-side first? The API is designed to analyze the entire image provided, not just a portion of it. – cthrash May 16 '18 at 23:03
  • I have to read a handwritten form, cropping and sending parts of image means approx. 50-60 API calls per form, this increases the time for processing. – Abhinav Mohan May 17 '18 at 04:49
  • The API will return bounding boxes of each text fragment, so unless I misunderstood you, can't you just make the one call? – cthrash May 17 '18 at 05:26
  • Yes, but another thing is that the accuracy is low when whole image is input, but if i crop small pieces and input them, accuracy increases. Also, bounding boxes (BB) change with every different image input. So instead of relying on BBs given in result, i want to send my own BBs as input. This way, i increase accuracy and also know the output for each BB i specified in input. – Abhinav Mohan May 17 '18 at 06:53
  • Also, its a waste of time to read all data on form with OCR. I want only hand written data to be read. That saves time. – Abhinav Mohan May 17 '18 at 07:00

0 Answers0