An Uncertainty Estimation Framework for Probabilistic Object Detection

Lyu, Zongyao; Gutierrez, Nolan B.; Beksi, William J.

doi:10.1109/case49439.2021.9551543

Search citation statements

Order By: Relevance

Paper Sections

Select...

Introduction1

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2022

2023

Publication Types

Select...

Other1

Article1

Relationship

Self Cite1

Independent1

Authors

Journals

Cited by 2 publications

(1 citation statement)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Within the realm of robotics, this motivates research questions such as how much trust can we put into the predictions of a DNN when misclassifications may have catastrophic consequences [2]? Methods have been proposed to mitigate the overconfidence problem by calibrating the predictive probabilities [3], or estimating the predictive uncertainties [4,5], for various vision-based objectives (e.g., image classification [6], semantic segmentation [7], and object detection [8,9]).…”

Section: Introductionmentioning

confidence: 99%

Evaluating Uncertainty Calibration for Open-Set Recognition

Lyu¹,

Gutierrez²,

Beksi³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Despite achieving enormous success in predictive accuracy for visual classification problems, deep neural networks (DNNs) suffer from providing overconfident probabilities on out-of-distribution (OOD) data. Yet, accurate uncertainty estimation is crucial for safe and reliable robot autonomy. In this paper, we evaluate popular calibration techniques for openset conditions in a way that is distinctly different from the conventional evaluation of calibration methods on OOD data. Our results show that closed-set DNN calibration approaches are much less effective for open-set recognition, which highlights the need to develop new DNN calibration methods to address this problem.

show abstract