On the Evaluation of RGB-D-based Categorical Pose and Shape Estimation

TL;DR

This paper critically evaluates current RGB-D-based pose and shape estimation methods, proposing new metrics, dataset annotations, and an evaluation toolbox to improve fair comparison and reveal biases towards upright objects.

Contribution

It introduces new evaluation metrics, provides updated annotations for the Redwood dataset, and offers an accessible toolbox for standardized assessment of pose and shape estimation methods.

Findings

Existing methods struggle with unconstrained orientations.

Methods are biased towards upright object poses.

The new evaluation toolbox facilitates fair comparison.

Abstract

Recently, various methods for 6D pose and shape estimation of objects have been proposed. Typically, these methods evaluate their pose estimation in terms of average precision, and reconstruction quality with chamfer distance. In this work we take a critical look at this predominant evaluation protocol including metrics and datasets. We propose a new set of metrics, contribute new annotations for the Redwood dataset and evaluate state-of-the-art methods in a fair comparison. We find that existing methods do not generalize well to unconstrained orientations, and are actually heavily biased towards objects being upright. We contribute an easy-to-use evaluation toolbox with well-defined metrics, method and dataset interfaces, which readily allows evaluation and comparison with various state-of-the-art approaches (see https://github.com/roym899/pose_and_shape_evaluation ).