Skip to content

(videoCalibration): speedup chessboard detection

Kevin GAUDET requested to merge speedup-chessboard-detection into dev

What does this MR do?

  • Add a scaling factor to the input to run the detection algorithm on a downscaled image.
  • Const-overload ::cvIO::moveToCV.
  • Add a new preference parameter to configure the scaling in ARCalibration.

How to test it?

It is advised to test this in Release mode.

  • Launch ARCalibration
  • Test the calibration activity
  • Try setting the scale in the chessboard parameters
  • Set to 1, 0.5 and 0.25 and notice the speed-up.

Some results

All measurements were done on my machine : Windows 10 Intel Core-i7 6700K, 32Gb RAM

For a full HD image (1920x1080):

Scale Average detection time
1 60ms
0.5 18ms
0.25 8ms

Associated Issues/Merge Requests

Edited by Kevin GAUDET

Merge request reports