I am implementing a standard VO algorithm, with a few changes, i.e. extract features, match the feature points, find the essential matrix and decompose to get the pose. Afte