A little question about track.hpp

[StraySparks]

Line 135-136 in track.hpp, I wonder if "const auto & feat = map_node_to_index[k]; const uint32_t & track_id = uf_tree.Find(feat.second);" make it more reasonable?
Because in line 188, the input of "uf_tree.Union()" is "map_node_to_index[k].second", the input of "uf_tree.Find()" should be the same, but "k" (line 135) is the index of "m_vec" in "class flat_pair_map". The code works well in line 135 because the value of "k" and "map_node_to_index[k].second" is the same when elements are already sorted after allFeatures.emplace() (type "set", line 74, 85), which means "map_node_to_index.sort()" actually did not work in line 100. If "map_node_to_index.sort()" worked, then the value of "k" and "map_node_to_index[k].second" will be different. The same question also happens in line 184.
I wonder if I've figured it out. Looking forward to your answers and thank you!

Best wishes!

[pmoulon]

Can you propose the change in a PR so we can investigate and clearly see the change you are proposing?

[StraySparks]

Can you propose the change in a PR so we can investigate and clearly see the change you are proposing?
I will try it later.

[StraySparks]

Another issue but not related to the above, line 54 in file "multiview/solver_resection_p3p_kneip.cpp" why using " ((P2-P1).cross(P3-Vec3(1,1,1))).norm() == 0" not "((P2-P1).cross(P3-P1))" to verify that world points are not colinear?

[pmoulon]

For this last one.. fell free to make the change locally and run the unit test. It seems indeed a bit strange.