I have reconstructed two frames using 5-point essential matrix, for sequential frames I am using PNP to get the camera pose before reconstructing the matches.
When I use video frames progressing through a corridor the results of PNP are terrible?
Is this a degenerate case?