0
$\begingroup$

I have been working with face recognition with neural networks. My dataset has images which fall short of pixels on one axis, as required by the chosen face alignment method. These short pixels may vary from 4 to 7% of pixels on that axis. Which ends up causing black borders on the images.

I would like to know if this is a common problem ? To what extent can it affect the accuracy ? How do people deal with it ?

Thanks.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Yes this is a common problem.

It shouldn't affect accuracy, as if the face is visible, the features will be picked up and recognized whatsoever.

Best way to deal with it is by using replicated borders, where instead of making a black border, the edge of the border would be replicated. Look up how to replicated borders in OpenCV to find out more.

Improve this answer
$\endgroup$

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.