Abstract:
Pose estimation is a fundamental problem in computer vision and is prevalent in numerous
applications such as object-detection, tracking etc. Generally, in pose estimation between two
camera views, we have 6 degrees of freedom - 3 rotational and 3 translational parameters. But
in some cases, we can enforce extra constraints such as a fixed axis of rotation. We can thus
frame our problem in terms of fewer parameters. In this report we discuss some of these
special cases and how we can perform pose estimation better for such cases.