unexpected large reprojection error returned by solvePnPRansac
hi,
is it an expected behaviour when solvePnPRansac returns this large error?
inlier 1 -> 270.973
inlier 2 -> 79.4406
inlier 9 -> 83.996
inlier 10 -> 44.9531
inlier 18 -> 126.671
inlier 22 -> 109.115
regardless I set error threshold is small ~15-30. with such large error, estimation result is mostly wrong. Am I missing some parameters?
I am using opencv 3.4.1 and run solvePnPRansac with default algorithm
I switch to SOLVEPNP_EPNP and for cases, the problem appears again. According to my understanding, RANSAC should exit when all conditions met (confidence, inlier rate, reprojection error). It looks like in this case, it exits when only one, two met or a weighted stop condition?
I try different algorithm than the default, SOLVEPNP_EPNP instead of default SOLVEPNP_ITERATIVE. This case, the reprojection error makes sense, bounded by input settings. Still, it means the default option is giving wrong results