Aleatoric Uncertainty from AI-based 6D Object Pose Predictors for Object-relative State Estimation

Thomas Jantos, Stephan Weiss, Jan Steinbrener · 2025

Read Aleatoric Uncertainty from AI-based 6D Object Pose Predictors for Object-relative State Estimation by Thomas Jantos, Stephan Weiss, Jan Steinbrener (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-01583.txt.

Description

Robots that navigate by camera need to know where an object is. They also need to know how unsure that reading is. Thomas Jantos, Stephan Weiss, and Jan Steinbrener extend deep learning pose predictors. These estimate an object's six-dimensional pose. The extension makes them report their own aleatoric uncertainty, the noise inherent in the data. Two small added networks handle the shift and rotation parts separately.

The add-ons sit apart from the original predictor. Training stays efficient, and the pre-trained model can remain frozen. The predicted pose and its uncertainty then feed an extended Kalman filter. They enter as a measurement with its matching noise matrix. The authors stress the low added computation. The approach suits object-relative navigation, where a robot steers by the objects it sees and weighs each reading by its trust.

Catalog record

CREATORThomas Jantos, Stephan Weiss, Jan Steinbrener
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-01583

Browse the full library · About the Tower of Babel