A simple yet effective baseline for 3d human pose estimation

TL;DR

This paper demonstrates that a simple deep learning approach can effectively lift 2d joint locations to 3d poses with low error, revealing that much of the current error in 3d pose estimation stems from visual analysis rather than the lifting process itself.

Contribution

The authors show that a straightforward deep feed-forward network can outperform complex methods in 3d pose lifting, and that using 2d detector outputs yields state-of-the-art results.

Findings

Simple lifting network outperforms previous methods by 30% on Human3.6M.

Using 2d detector outputs achieves state-of-the-art 3d pose estimation.

Majority of errors in 3d pose estimation originate from 2d visual understanding.

Abstract

Following the success of deep convolutional networks, state-of-the-art methods for 3d human pose estimation have focused on deep end-to-end systems that predict 3d joint locations given raw image pixels. Despite their excellent performance, it is often not easy to understand whether their remaining error stems from a limited 2d pose (visual) understanding, or from a failure to map 2d poses into 3-dimensional positions. With the goal of understanding these sources of error, we set out to build a system that given 2d joint locations predicts 3d positions. Much to our surprise, we have found that, with current technology, "lifting" ground truth 2d joint locations to 3d space is a task that can be solved with a remarkably low error rate: a relatively simple deep feed-forward network outperforms the best reported result by about 30\% on Human3.6M, the largest publicly available 3d pose estimation benchmark. Furthermore, training our system on the output of an off-the-shelf state-of-the-art 2d detector (\ie, using images as input) yields state of the art results -- this includes an array of systems that have been trained end-to-end specifically for this task. Our results indicate that a large portion of the error of modern deep 3d pose estimation systems stems from their visual analysis, and suggests directions to further advance the state of the art in 3d human pose estimation.