Joint Training of a Convolutional Network and a Graphical Model for\n Human Pose Estimation

Abstract

This paper proposes a new hybrid architecture that consists of a deep\nConvolutional Network and a Markov Random Field. We show how this architecture\nis successfully applied to the challenging problem of articulated human pose\nestimation in monocular images. The architecture can exploit structural domain\nconstraints such as geometric relationships between body joint locations. We\nshow that joint training of these two model paradigms improves performance and\nallows us to significantly outperform existing state-of-the-art techniques.\n