1
$\begingroup$

I understand (I think) why in object detection, the result is a rectangle:

it is a simple shape that can be defined by 4 variables (2 pairs coords of opposite corners or 1 pair of coords + width and height)

So more complicated shape might require more parameters which could complicate things. But for example, what if a circle was used? There would just be 3 parameters, 1 pair of coordinates of the center + the radius. Is there is something obvious I am missing?

And still regarding the bounding box, I wonder what would happen if a 5-th parameter was added that would describe the angle of the bbox. For example, consider the iPhone on this image:

enter image description here

I might be thinking about it wrong, but in my head, the network could have an easier time "understanding" that we want it to detect a rotated bbox that alignes with the actual iPhone than an axes-aligned bbox. For a human, it's also easier to draw the rotated bounding box (and i would argue, more intuitive) than the axes-aligned, isn't it?

And regardless of whether it would be easier for the network, the rotated bounding box would be a more precise result of the detection.

$\endgroup$

1 Answer 1

0
$\begingroup$

There is some work on this, came across this blog.
https://developer.nvidia.com/blog/detecting-rotated-objects-using-the-odtk/

$\endgroup$
1
  • 1
    $\begingroup$ Your answer could be improved with additional supporting information. Please edit to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers in the help center. $\endgroup$
    – Community Bot
    Mar 23, 2022 at 15:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.