ECCV 2018 Unsupervised Geometry-Aware Representation Learning for 3D Human Pose Estimation
Modern 3D human pose estimation techniques rely on deep networks, which require large amounts of training data.
In this paper, we propose to overcome this problem by learning a geometry-aware body representation from multi-view images without human annotations. Because this representation encodes 3D geometry, using it in a semi-supervised setting makes it easier to learn a mapping from it to 3D human pose. As evidenced by our experiments, our approach improves over other semi-supervised methods while using as little as 1% of the labeled data.
https://arxiv.org/abs/1804.01110
By Helge Rhodin, Mathieu Salzmann and Pascal Fua
