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

Although deep learning-based segmentation models have achieved impressiveperformance on public benchmarks, generalizing well to unseen environmentsremains a major challenge. To improve the model's generalization ability to thenew domain during evaluation, the test-time training (TTT) is a challengingparadigm that adapts the source-pretrained model in an online fashion. Earlyefforts on TTT mainly focus on the image classification task. Directlyextending these methods to semantic segmentation easily experiences unstableadaption due to segmentation's inherent characteristics, such as extreme classimbalance and complex decision spaces. To stabilize the adaptation process, weintroduce contrastive loss (CL), known for its capability to learn robust andgeneralized representations. Nevertheless, the traditional CL operates in therepresentation space and cannot directly enhance predictions. In this paper, weresolve this limitation by adapting the CL to the output space, employing ahigh temperature, and simplifying the formulation, resulting in astraightforward yet effective loss function called Output Contrastive Loss(OCL). Our comprehensive experiments validate the efficacy of our approachacross diverse evaluation scenarios. Notably, our method excels even whenapplied to models initially pre-trained using domain adaptation methods on testdomain data, showcasing its resilience and adaptability.\footnote{Code and moreinformation could be found at~ \url{https://github.com/dazhangyu123/OCL}}

Test-Time Training for Semantic Segmentation with Output Contrastive Loss