Adjoint Rigid Transform Network: Task-conditioned Alignment of 3D Shapes
This addresses the need for automated alignment of 3D data from diverse sources, which is crucial for improving accuracy in various 3D tasks, though it appears incremental as it builds on existing neural modules.
The paper tackles the problem of performance drops in 3D learning methods due to misaligned data by proposing the Adjoint Rigid Transform (ART) Network, which learns to rotate shapes to a canonical orientation using self-supervision, resulting in a notable boost in performance for tasks like shape reconstruction and interpolation.
Most learning methods for 3D data (point clouds, meshes) suffer significant performance drops when the data is not carefully aligned to a canonical orientation. Aligning real world 3D data collected from different sources is non-trivial and requires manual intervention. In this paper, we propose the Adjoint Rigid Transform (ART) Network, a neural module which can be integrated with a variety of 3D networks to significantly boost their performance. ART learns to rotate input shapes to a learned canonical orientation, which is crucial for a lot of tasks such as shape reconstruction, interpolation, non-rigid registration, and latent disentanglement. ART achieves this with self-supervision and a rotation equivariance constraint on predicted rotations. The remarkable result is that with only self-supervision, ART facilitates learning a unique canonical orientation for both rigid and nonrigid shapes, which leads to a notable boost in performance of aforementioned tasks. We will release our code and pre-trained models for further research.