Symmetry-Robust 3D Orientation Estimation

Christopher Scarvelis, David Benhaim, Paul Zhang

TL;DR: We introduce a state-of-the-art method for 3D orientation estimation.

Abstract: Orientation estimation is a fundamental task in 3D shape analysis which consists of estimating a shape's orientation axes: its side-, up-, and front-axes. Using this data, one can rotate a shape into canonical orientation, where its orientation axes are aligned with the coordinate axes. Developing an orientation algorithm that reliably estimates complete orientations of general shapes remains an open problem. We introduce a two-stage orientation pipeline that achieves state of the art performance on up-axis estimation and further demonstrate its efficacy on full-orientation estimation, where one seeks all three orientation axes. Unlike previous work, we train and evaluate our method on all of Shapenet rather than a subset of classes. We motivate our engineering contributions by theory describing fundamental obstacles to orientation estimation for rotationally-symmetric shapes, and show how our method avoids these obstacles.

Lay Summary: Estimating the orientation of 3D shapes – figuring out which way is up, front, and side – is key to many tasks in computer graphics and vision. This helps standardize how shapes are viewed and used by aligning them with a common frame of reference. However, building a method that works reliably for all kinds of shapes, especially those with rotational symmetries, is still a challenge. In this work, we present a two-step method that sets a new standard for estimating a shape’s upright direction and shows strong results for estimating its full orientation, which includes its front and side-facing directions. Unlike earlier approaches that were tested only on a few shape categories, our method is trained and tested on the entire ShapeNet dataset. We also offer a theoretical explanation for why shapes with rotational symmetries are harder to orient, and we specifically design our approach to overcome these challenges.

Primary Area: Applications->Computer Vision

Keywords: 3d orientation, shape analysis, 3d deep learning, geometric deep learning

Submission Number: 7672