2015-01-05 06:46:46 -0600 | received badge | ● Student (source) |
2015-01-04 12:59:57 -0600 | asked a question | superresolution performance Hi I am trying out the superresolution samples on a simple input video (1280x720, 23.14 fps) with a duration of only 1 second and it is taking an excruciating amount of time to process (on the scale of minutes, instead of seconds) I changed the number of iterations down to 1 (instead of the default 180) which resulted in a somewhat better performance but still on the scale of minutes. I had not compiled opencv for CUDA or OCL but I do not expect the net gain to be of such that the whole thing runs in seconds. Question to those with experience with superresolution: Is this the expected behavior? If so perhaps it should be added to the documentation? If a much faster performance is expected, could you kindly provide with some hints? best regards |
2014-08-29 04:50:59 -0600 | commented answer | how many images for camera calibration? of course according to Zhang's paper it seems more samples the better the results. In fact according to the data presented there (if I remember correctly) there are dramatic improvements with 5+ images. However in many tests I got a much lower reprojection error rate (~.2 or .3) when I used 5-10 images versus 20-30. In my view it ought be possible for calibration in a controlled environment to use images taken from a minimal set of fixed orientations to get a reliable result. Instead of the adhoc approach currently. That needs re investigating the mathematical derivation. As far as near or far camera positions relative to the chess board plane: I got my best results (better undistortion) when all the pictures were taken with the camera closest to the plane. Thanks. |
2014-08-28 07:06:37 -0600 | asked a question | how many images for camera calibration? Hi, I finally manged to get camera calibration working with relatively reliable results (using a set of 10 standard chessboard images). However the choice of the orientation of the chess board seems rather ad hoc. Does anyone in this forum know of a minimal set of standard orientations (e.g. n=(0, 0, 1), (0, -1, 1), (-1, 0,1) ...) which would result in a stable camera calibration? In a controlled environment this ought be possible? In the provided examples it seems like people hold the board in front of the camera and hope for the best. Thank you |