Reconstructing 3D human poses from monocular image

This project aims to reconstruct 3D human poses from a small number of 2D points obtained from a calibrated monocular image. With homography transformation and camera calibration, we can limit the solution space and then use some human pose constraints to choose the best possible solution. We believe that this methodology will be effective for our future works on musculoskeletal disorder assessment from 3D poses of patients walking or running on a treadmill.