I have trained a CNN network to detect a circle and approximate its centre and radius in an image. What I want to do now is detect the centre and radius of all the circles if there are multiple circles present in an image.

How do I proceed to go on about it? Do I have to make changes to my dataset to be able to do so? I tried to look at different architectures that do multiple object detection, but I couldn't understand what changes I could make to my architecture.

  • 1
    $\begingroup$ I won't be able to give you the exact answer but I will direct you to anchor boxes in object detection. The idea is to create multiple bounding boxes with different shapes centered around each pixel (or pixel groups) such that you can optimize the network parameters based on loss functions such as Intersection-over-Union (IoU) over many ground truth bounding boxes instead of just one. Hope this helps :) $\endgroup$
    – Arun Das
    Jun 16 at 15:06
  • $\begingroup$ It does help a little, it would be great if you could share the link of an example for the same so that I can understand it better. $\endgroup$
    – Ravish Jha
    Jun 16 at 16:30
  • $\begingroup$ One of the best content for anchor boxes is here: d2l.ai/chapter_computer-vision/anchor.html You can also look at ronny.rest/tutorials/module/localization_001/iou and stackoverflow.com/questions/25349178/… to read about IoU. $\endgroup$
    – Arun Das
    Jun 17 at 2:41

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.