English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Motivated by the increasing popularity of transformers in computer vision, inrecent times there has been a rapid development of novel architectures. Whilein-domain performance follows a constant, upward trend, properties likerobustness or uncertainty estimation are less explored -leaving doubts aboutadvances in model reliability. Studies along these axes exist, but they aremainly limited to classification models. In contrast, we carry out a study onsemantic segmentation, a relevant task for many real-world applications wheremodel reliability is paramount. We analyze a broad variety of models, spanningfrom older ResNet-based architectures to novel transformers and assess theirreliability based on four metrics: robustness, calibration, misclassificationdetection and out-of-distribution (OOD) detection. We find that while recentmodels are significantly more robust, they are not overall more reliable interms of uncertainty estimation. We further explore methods that can come tothe rescue and show that improving calibration can also help with otheruncertainty metrics such as misclassification or OOD detection. This is thefirst study on modern segmentation models focused on both robustness anduncertainty estimation and we hope it will help practitioners and researchersinterested in this fundamental vision task. Code available athttps://github.com/naver/relis.

Reliability in Semantic Segmentation: Are We on the Right Track?