We introduce AR Poser: a framework for posing with, or as a digital character. In this paper, we describe our first contribution to AR Poser: a technique for digital characters to recognize and automatically reproduce the same pose as a person in a picture (using only RGB information from a mobile device). 3D human pose estimation from RGB is an under-constrained and ambiguous problem that remains today an active field of study. Instead of addressing the general case of human pose estimation, we propose a solution that can be tailored to a specific scenario—such as entertainment poses for AR selfies. At the heart of our solution is a set of predefined poses (selfie poses) utilized to reduce ambiguities. In a nutshell, our method consists of two reliable steps: we first perform 2D pose estimation, and then perform a projection onto the 3D subspace to find the closest matching 3D pose. With our method, we are able to automatically create augmented reality selfies for a variety of different poses.
The documents contained in these directories are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author’s copyright. These works may not be reposted without the explicit permission of the copyright holder.