Uncertainty Quantification Using Query-Based Object Detectors
Abstract
Recently, a new paradigm of query-based object detection has gained popularity. In this paper, we study the problem of quantifying the uncertainty in the predictions of these models that derive from model uncertainty. Such uncertainty quantification is vital for many high-stakes applications that need to avoid making overconfident errors. We focus on quantifying multiple aspects of detection uncertainty based on a deep ensembles representation. We perform extensive experiments on two representative models in this space: DETR and AdaMixer. We show that deep ensembles of these query-based detectors result in improved performance with respect to three types of uncertainty: location uncertainty, class uncertainty, and objectness uncertainty (Code available at: https://github.com/colinski/uq-query-object-detectors ).
Cite
Text
Vadera et al. "Uncertainty Quantification Using Query-Based Object Detectors." European Conference on Computer Vision Workshops, 2022. doi:10.1007/978-3-031-25085-9_5Markdown
[Vadera et al. "Uncertainty Quantification Using Query-Based Object Detectors." European Conference on Computer Vision Workshops, 2022.](https://mlanthology.org/eccvw/2022/vadera2022eccvw-uncertainty/) doi:10.1007/978-3-031-25085-9_5BibTeX
@inproceedings{vadera2022eccvw-uncertainty,
title = {{Uncertainty Quantification Using Query-Based Object Detectors}},
author = {Vadera, Meet P. and Samplawski, Colin and Marlin, Benjamin M.},
booktitle = {European Conference on Computer Vision Workshops},
year = {2022},
pages = {78-93},
doi = {10.1007/978-3-031-25085-9_5},
url = {https://mlanthology.org/eccvw/2022/vadera2022eccvw-uncertainty/}
}