Human pose estimation is an image processing/computer vision task that predicts different keypoints' (joints/landmarks—for example, elbows, knees, neck, shoulder, hips, and chest) locations in the human skeleton, representing the pose (orientation) of a human (sets of coordinates are connected to find a person's overall pose). A limb/pair is defined by a valid connection between two parts; some combinations of two parts may not form valid pairs.
Multiperson pose estimation is harder than its single-person counterpart, as both the number of people in an image along with the locations are not known. As a bottom-up approach to solve this problem, all parts of the image for all the featured people are first detected and then those parts corresponding to individuals are associated/grouped. A popular bottom-up approach...
