We present a method for learning an embedding that places images of humans in similar poses nearby. This embedding can be used as a direct method of comparing images based on human pose, avoiding potential challenges of estimating body joint positions. Pose embedding learning is formulated under a triplet-based distance criterion. A deep architecture is used to allow learning of a representation capable of making distinctions between different poses. Experiments on human pose matching and retrieval from video data demonstrate the potential of the method.
@article{arxiv.1507.00302,
title = {Pose Embeddings: A Deep Architecture for Learning to Match Human Poses},
author = {Greg Mori and Caroline Pantofaru and Nisarg Kothari and Thomas Leung and George Toderici and Alexander Toshev and Weilong Yang},
journal= {arXiv preprint arXiv:1507.00302},
year = {2015}
}