Towards Symmetry-sensitive Pose Estimation: A Rotation Representation for Symmetric Object Classes

TL;DR

This paper introduces SARR, a novel rotation representation tailored for symmetric objects in pose estimation, enabling standard CNNs to produce symmetry-aware 3D orientations without requiring 3D models.

Contribution

The authors propose SARR, a continuous and unique rotation representation for symmetric objects, improving pose estimation accuracy and symmetry handling in deep learning models.

Findings

SARR outperforms state-of-the-art methods on T-LESS and ITODD datasets.

Networks trained with SARR achieve higher symmetry-sensitive accuracy.

The method works with depth and texture-less RGB images without 3D models.

Abstract

Symmetric objects are common in daily life and industry, yet their inherent orientation ambiguities that impede the training of deep learning networks for pose estimation are rarely discussed in the literature. To cope with these ambiguities, existing solutions typically require the design of specific loss functions and network architectures or resort to symmetry-invariant evaluation metrics. In contrast, we focus on the numeric representation of the rotation itself, modifying trigonometric identities with the degrees of symmetry derived from the objects' shapes. We use our representation, SARR, to obtain canonic (symmetry-resolved) poses for the symmetric objects in two popular 6D pose estimation datasets, T-LESS and ITODD, where SARR is unique and continuous w.r.t. the visual appearance. This allows us to use a standard CNN for 3D orientation estimation whose performance is evaluated with the symmetry-sensitive cosine distance $\text{AR}_{\text{C}}$. Our networks outperform the state of the art using $\text{AR}_{\text{C}}$ and achieve satisfactory performance when using conventional symmetry-invariant measures. Our method does not require any 3D models but only depth, or, as part of an additional experiment, texture-less RGB/grayscale images as input. We also show that networks trained on SARR outperform the same networks trained on rotation matrices, Euler angles, quaternions, standard trigonometrics or the recently popular 6d representation -- even in inference scenarios where no prior knowledge of the objects' symmetry properties is available. Code and a visualization toolkit are available at https://github.com/akriegler/SARR .