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

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

Automated summary

Human pose estimation is a crucial research field with important applications in robotics, entertainment, health, and sports sciences. Recent advancements in convolutional networks have led to significant improvements in 2D pose estimation and reduced the average error in modern 3D markerless motion capture techniques. However, the increasing number of methods in this field has made it challenging to make informed choices. Researchers have proposed a taxonomy based on accuracy, speed, and robustness to categorize methods and provide guidance for future research.