A Laplace-inspired Distribution on SO(3) for Probabilistic Rotation Estimation
This work addresses a domain-specific challenge in computer vision for applications like robotics or augmented reality, offering incremental improvements in probabilistic modeling for rotation estimation.
The paper tackles the problem of probabilistic 3D rotation estimation from RGB images by proposing a Rotation Laplace distribution on SO(3) to improve robustness to outliers and convergence, achieving state-of-the-art performance in rotation regression tasks.
Estimating the 3DoF rotation from a single RGB image is an important yet challenging problem. Probabilistic rotation regression has raised more and more attention with the benefit of expressing uncertainty information along with the prediction. Though modeling noise using Gaussian-resembling Bingham distribution and matrix Fisher distribution is natural, they are shown to be sensitive to outliers for the nature of quadratic punishment to deviations. In this paper, we draw inspiration from multivariate Laplace distribution and propose a novel Rotation Laplace distribution on SO(3). Rotation Laplace distribution is robust to the disturbance of outliers and enforces much gradient to the low-error region, resulting in a better convergence. Our extensive experiments show that our proposed distribution achieves state-of-the-art performance for rotation regression tasks over both probabilistic and non-probabilistic baselines. Our project page is at https://pku-epic.github.io/RotationLaplace.