This is the sample document -> enter image description here

I want to extract questions along with the options. There are other question papers as which have questions with diagrams in them. I want to be able to extract them as well.

End goal is to store them in a question bank.

I am planning to run an image segmentation algorithm to extract the questions along with their options. Will that be feasible? Suggestions are more than welcome.

  • $\begingroup$ you need to extract the text, or only the part of the image (as image) that represents a single question? $\endgroup$ – Nikos M. Jul 26 '20 at 13:56
  • $\begingroup$ Part of the image. Because I need the figures as well corresponding to that particular question. $\endgroup$ – Faraz Gerrard Jamal Jul 26 '20 at 17:57
  • 2
    $\begingroup$ well the problem has certain visual symmetries and hints. For example each question is marked by a question number which is at the left of the image at a certain distance from left edge. The end of the question is marked by the next question number or the bottom edge of the image. I think a custom computer vision algorithm based on those symmetries is more than enough and very reliable, than an ANN which would need many labelled samples $\endgroup$ – Nikos M. Jul 27 '20 at 7:21
  • $\begingroup$ @NikosM. what if there are >=2 columns of questions in a question paper, along with diagrams? $\endgroup$ – Faraz Gerrard Jamal Jul 29 '20 at 4:54
  • $\begingroup$ Similar visual cues can be used, in fact one cannot accept arbitrary images (whatever the segmentation method used), their visual architecture should be known beforehand, in fact it can even be a parameter (even hardcoded) so one can run the computer vision algorithm and specify which image format is being used each time $\endgroup$ – Nikos M. Jul 29 '20 at 6:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.