The field of computer vision is moving towards incorporating uncertainty modeling and rotation equivariance into its frameworks. This is evident in the development of novel parameterizations that capture intrinsic correlations among variables, allowing for better trade-offs between modeling complexity and computational efficiency. Additionally, there is a growing interest in leveraging quaternion algebra for rotation equivariant image classification and object detection, which preserves geometric properties while enabling efficient implementation. Noteworthy papers include:

• Learning Correlation-aware Aleatoric Uncertainty for 3D Hand Pose Estimation, which introduces aleatoric uncertainty modeling into 3D hand pose estimation frameworks.
• Quaternion Approximation Networks for Enhanced Image Classification and Oriented Object Detection, which achieves higher accuracy with fewer parameters and faster convergence compared to existing convolution and quaternion-based models.