3D Hand Pose Estimation in the Wild via Graph Refinement under Adversarial Learning.

This paper addresses the problem of 3D hand pose estimation from a monocular RGB image. While previous methods have shown their success, the structure of hands has not been exploited explicitly, which is critical in pose estimation. To this end, we propose a hand-model regularized graph refinement paradigm under an adversarial learning framework, aiming to explicitly capture structural inter-dependencies of hand joints for the learning of intrinsic patterns. We estimate an initial hand pose from a parametric hand model as a prior of hand structure, and refine the structure by learning the deformation of the prior pose via residual graph convolution. To optimize the hand structure further, we propose two bone-constrained loss functions, which characterize the morphable structure of hand poses explicitly. Also, we introduce an adversarial learning framework with a multi-source discriminator to capture structural features, which imposes the constraints onto the distribution of generated 3D hand poses for anthropomorphically valid hand poses. Extensive experiments demonstrate that our model sets the new state-of-the-art in 3D hand pose estimation from a monocular image on standard benchmarks.