3 ms·
I'm not sure how good your algorithm is if it can only work in a specific orientation. A slightly tilted image can already cause problems for such an algorithm
by jeroenhd 7y ago
I'm not sure how good your algorithm is if it can only work in a specific orientation. A slightly tilted image can already cause problems for such an algorithm and many people have a hard time getting a picture with only a few degrees of rotation.
It would probably be better to deal with this in an elegant way, e.g. set up the algorithm to work regardless of orientation. This seems like a (mostly) solved problem: https://d4nst.github.io/2017/01/12/image-orientation/ https://d4nst.github.io/2017/01/12/image-orientation/
- H8crilA 7y agoForcing rotation in the training phase would be a good way to regularize the network. But of course that will lower the accuracy.
- gugagore 7y agoYou might not want rotation invariance in your algorithm. Our own perceptual system is not rotation-invariant, most obviously for faces in this illusion: https://en.wikipedia.org/wiki/Thatcher_effect https://en.wikipedia.org/wiki/Thatcher_effect
- cookingrobot 7y agoThat’s not an option if you’re trying to detect left-arrows vs right-arrows, on street signs for ex.
- jeroenhd 7y agoDepends. Based on the linked article, the picture can be completely upside down if it was taken in landscape mode. The article I linked is capable of rotating it upright regardless. After rotating to the right orientation, left and right are suddenly perfectly workable. Of course using exif data for such rotations is easier, but a tilted picture of a tilted sign can create a lot of tilt that human vision copes with fine but an orientation dependent network cannot.