Closed coder12333444 closed 2 years ago
Each C_i, together with the rotation matrix R, defines a pose. For this pose, we count the number of 2D-3D matches (created in a previous step) that are inliers to this pose by projecting the 3D points into the images and measuring the reprojection error compared to the corresponding 2D coordinates.
The corresponding part of the code is this: https://github.com/tsattler/meshloc_release/blob/main/src/pose_estimation.cc#L640-L673
@v-pnk @tsattler Hi,I'm confused about the underlined part, since Ci is a sampling point based on C (virtual), how do I go about counting the number of inliers points in 2d-3D?