I am using the Aruco library in openCV(c++) to detect a big number of very small barcodes(80-120), using a camera to take frames in very high resolution and running Aruco detection algorithms on the frames. FPS is 18.
If relevent, The barcodes format is the following:
Each barcode gets detected by its black/white pixels. its value connected to some id the barcode repesent.
On one computer, the detection is really bad. Like 30% detection success rate. On another computer, with stronger CPU and GPU - the detection success rate is like 90%+.
Both computers use nivida GPU. All other variables like light, camera settings and so on are the same.
My question is more general and not specific to those computers, as i assume that every detection algorithm is effected by the hardware:
Why better GPU and/or CPU effect detection success rate? as opposed to latency which is caused by slower hardware, i don't understand why the detection quality itself is effected. is information get lost on weaker computers? is the math operations speed effect detection? if so,why?
On the slower computer, i tried everything to improve detection. Playing with light, camera settings, focus.. nothing helps. switching to new computer fixed everything.
Thank you.
