On the Role of Rotation Equivariance in Monocular 2D-to-3D Human Pose Lifting

Estimating 3D from 2D is one of the central tasks in computer vision. In this work, we consider the monocular setting, i.e. single-view input, for 3D human pose estimation (HPE), where the goal is to predict a 3D point set of human skeletal joints from a single 2D image, typically via 2D keypoint detection followed by 2D-to-3D lifting. Despite their success, we find that current lifting models exhibit strong performance degradation under rotations. We address this by considering different approaches to incorporating rotation equivariance, including explicit equivariant architectures and standard models. Utilising common HPE benchmarks, we demonstrate that rotation equivariance can be effectively learned via rotation-based data augmentation applied jointly to input and output poses. This significantly improves robustness to rotations and, in this setting, outperforms methods that are fully equivariant by design, while maintaining a lower computational cost.