A review of 3D human pose estimation algorithms for markerless motion capture

Human pose estimation is a very active research field, stimulated by its\nimportant applications in robotics, entertainment or health and sports\nsciences, among others. Advances in convolutional networks triggered noticeable\nimprovements in 2D pose estimation, leading modern 3D markerless motion capture\ntechniques to an average error per joint of 20 mm. However, with the\nproliferation of methods, it is becoming increasingly difficult to make an\ninformed choice. Here, we review the leading human pose estimation methods of\nthe past five years, focusing on metrics, benchmarks and method structures. We\npropose a taxonomy based on accuracy, speed and robustness that we use to\nclassify de methods and derive directions for future research.\n