To do

Rectified frame not coming properly and Why is frame getting transposed when being rectified? - see cv2.stereorectify
pose estimation error - add fix_u in tf version
uncertainty zeros in den
Zero det hessian and other random ass zeros
Normalise depth and deal with zero disparity
not finding huber norm, just squaring right now
x and y interchange while generating point cloud and other places
Initialize pose to prev pose in pose est
see cv2.correct matches,cv2.reprojectimageto3d  - check out that whole page
remove all .numpy()
Show only certain points
Use pangolin
Semantic label seg
instead of fix_u, chuck out the point

# Possible error
I to F in keyframe class