Gaussian Latent Representations for Uncertainty Estimation using Mahalanobis Distance in Deep Classifiers

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21730%2F23%3A00374356" target="_blank" >RIV/68407700:21730/23:00374356 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1109/ICCVW60793.2023.00483" target="_blank" >https://doi.org/10.1109/ICCVW60793.2023.00483</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/ICCVW60793.2023.00483" target="_blank" >10.1109/ICCVW60793.2023.00483</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Gaussian Latent Representations for Uncertainty Estimation using Mahalanobis Distance in Deep Classifiers

Popis výsledku v původním jazyce

Recent works show that the data distribution in a network's latent space is useful for estimating classification uncertainty and detecting Out-Of-Distribution (OOD) samples. To obtain a well-regularized latent space that is conducive for uncertainty estimation, existing methods bring in significant changes to model architectures and training procedures. In this paper, we present a lightweight and high-performance regularization method for Mahalanobis distance (MD)-based uncertainty prediction, and that requires minimal changes to the network's architecture. To derive Gaussian latent representation favourable for MD calculation, we introduce a self-supervised representation learning method that separates in-class representations into multiple Gaussians. Classes with non-Gaussian representations are automatically identified and dynamically clustered into multiple new classes that are approximately Gaussian. Evaluation on standard OOD benchmarks shows that our method achieves state-of-the-art results on OOD detection and is very competitive on predictive probability calibration. Finally, we show the applicability of our method to a real-life computer vision use case on microorganism classification.

Název v anglickém jazyce

Gaussian Latent Representations for Uncertainty Estimation using Mahalanobis Distance in Deep Classifiers

Popis výsledku anglicky

Recent works show that the data distribution in a network's latent space is useful for estimating classification uncertainty and detecting Out-Of-Distribution (OOD) samples. To obtain a well-regularized latent space that is conducive for uncertainty estimation, existing methods bring in significant changes to model architectures and training procedures. In this paper, we present a lightweight and high-performance regularization method for Mahalanobis distance (MD)-based uncertainty prediction, and that requires minimal changes to the network's architecture. To derive Gaussian latent representation favourable for MD calculation, we introduce a self-supervised representation learning method that separates in-class representations into multiple Gaussians. Classes with non-Gaussian representations are automatically identified and dynamically clustered into multiple new classes that are approximately Gaussian. Evaluation on standard OOD benchmarks shows that our method achieves state-of-the-art results on OOD detection and is very competitive on predictive probability calibration. Finally, we show the applicability of our method to a real-life computer vision use case on microorganism classification.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

ICCVW2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

ISBN

979-8-3503-0744-3

ISSN

2473-9936

e-ISSN

2473-9944

Počet stran výsledku

10

Strana od-do

4490-4499

Název nakladatele

IEEE

Místo vydání

Anchorage, Alaska

Místo konání akce

Paris

Datum konání akce

2. 10. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001156680304062