Absolute Human Pose Estimation with Depth Prediction Network
This addresses the limitation of existing methods in handling multiple people, though it appears incremental as it builds on prior work in absolute coordinate prediction.
The paper tackles the problem of 3D human pose estimation for multiple interacting people by proposing a neural network that predicts joints in a camera-centered coordinate system in a single step, achieving state-of-the-art results on the MuPoTS-3D dataset.
The common approach to 3D human pose estimation is predicting the body joint coordinates relative to the hip. This works well for a single person but is insufficient in the case of multiple interacting people. Methods predicting absolute coordinates first estimate a root-relative pose then calculate the translation via a secondary optimization task. We propose a neural network that predicts joints in a camera centered coordinate system instead of a root-relative one. Unlike previous methods, our network works in a single step without any post-processing. Our network beats previous methods on the MuPoTS-3D dataset and achieves state-of-the-art results.