Introspective Robot Perception using Smoothed Predictions from Bayesian Neural Networks

TL;DR

This paper enhances robot perception by improving uncertainty estimation in object classification using Bayesian Neural Networks, and demonstrates benefits in domain adaptation and contextual reasoning in robotic applications.

Contribution

It introduces the use of Concrete Dropout and Laplace Approximation for better uncertainty estimates in BNNs, applied to robot perception tasks.

Findings

Improved uncertainty estimates lead to better object classification performance.

Uncertainty-based domain adaptation reduces manual annotation efforts.

Contextual information integration enhances perception accuracy.

Abstract

This work focuses on improving uncertainty estimation in the field of object classification from RGB images and demonstrates its benefits in two robotic applications. We employ a (BNN), and evaluate two practical inference techniques to obtain better uncertainty estimates, namely Concrete Dropout (CDP) and Kronecker-factored Laplace Approximation (LAP). We show a performance increase using more reliable uncertainty estimates as unary potentials within a Conditional Random Field (CRF), which is able to incorporate contextual information as well. Furthermore, the obtained uncertainties are exploited to achieve domain adaptation in a semi-supervised manner, which requires less manual efforts in annotating data. We evaluate our approach on two public benchmark datasets that are relevant for robot perception tasks.