Open gcarlin8 opened 4 years ago
Since you manually pick 3D/2D correspondences here, you can re-project the picked 3D points to the image after you have the extrinsics estimate and compute the pixel error with the 2D points you initially picked. Also have a look at OpenCV's PnP methods documentation which lets you set the inlier count, reprojection error thresholds, confidence, and other optimization criteria.
Hello heethesh,
Is there anyway to get the information about the point error/pixel error or any other error metrics