On the Continuity of Rotation Representations in Neural Networks

Abstract

In neural networks, it is often desirable to work with various representations of the same space. For example, 3D rotations can be represented with quaternions or Euler angles. In this paper, we advance a definition of a continuous representation, which can be helpful for training deep neural networks. We relate this to topological concepts such as homeomorphism and embedding. We then investigate what are continuous and discontinuous representations for 2D, 3D, and n-dimensional rotations. We demonstrate that for 3D rotations, all representations are discontinuous in the real Euclidean spaces of four or fewer dimensions. Thus, widely used representations such as quaternions and Euler angles are discontinuous and difficult for neural networks to learn. We show that the 3D rotations have continuous representations in 5D and 6D, which are more suitable for learning. We also present continuous representations for the general case of the n-dimensional rotation group SO(n). While our main focus is on rotations, we also show that our constructions apply to other groups such as the orthogonal group and similarity transforms. We finally present empirical results, which show that our continuous rotation representations outperform discontinuous ones for several practical problems in graphics and vision, including a simple autoencoder sanity test, a rotation estimator for 3D point clouds, and an inverse kinematics solver for 3D human poses.

References in corpus (4)

• DeMoN: Depth and Motion Network for Learning Monocular Stereo
• Deep-6DPose: Recovering 6D Object Pose from a Single RGB Image
• Explorations in Homeomorphic Variational Auto-Encoding
• Occlusion Resistant Object Rotation Regression from Point Cloud Segments

Cited by in corpus (15)

• DISN: Deep Implicit Surface Network for High-quality Single-view 3D Reconstruction
• IMUPoser: Full-Body Pose Estimation using IMUs in Phones, Watches, and Earbuds
• Radar Voxel Fusion for 3D Object Detection
• S4G: Amodal Single-view Single-Shot SE(3) Grasp Detection in Cluttered Scenes
• TransPose: Real-time 3D Human Translation and Pose Estimation with Six Inertial Sensors
• A Smooth Representation of Belief over SO(3) for Deep Rotation Learning with Uncertainty
• Anisotropic molecular coarse-graining by force and torque matching with neural networks
• Delving Deep into Pixel Alignment Feature for Accurate Multi-view Human Mesh Recovery
• Parametric Surface Constrained Upsampler Network for Point Cloud
• Pose Representations for Deep Skeletal Animation
• Monocular, One-stage, Regression of Multiple 3D People
• Learning a Disentangled Embedding for Monocular 3D Shape Retrieval and Pose Estimation
• Towards Using Active Learning Methods for Human-Seat Interactions To Generate Realistic Occupant Motion
• Unsupervised Dense Deformation Embedding Network for Template-Free Shape Correspondence
• Learning Realistic Human Reposing using Cyclic Self-Supervision with 3D Shape, Pose, and Appearance Consistency