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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Cross-modal Unsupervised Domain Adaptation (UDA) aims to exploit thecomplementarity of 2D-3D data to overcome the lack of annotation in a newdomain. However, UDA methods rely on access to the target domain duringtraining, meaning the trained model only works in a specific target domain. Inlight of this, we propose cross-modal learning under bird's-eye view for DomainGeneralization (DG) of 3D semantic segmentation, called BEV-DG. DG is morechallenging because the model cannot access the target domain during training,meaning it needs to rely on cross-modal learning to alleviate the domain gap.Since 3D semantic segmentation requires the classification of each point,existing cross-modal learning is directly conducted point-to-point, which issensitive to the misalignment in projections between pixels and points. To thisend, our approach aims to optimize domain-irrelevant representation modelingwith the aid of cross-modal learning under bird's-eye view. We proposeBEV-based Area-to-area Fusion (BAF) to conduct cross-modal learning underbird's-eye view, which has a higher fault tolerance for point-levelmisalignment. Furthermore, to model domain-irrelevant representations, wepropose BEV-driven Domain Contrastive Learning (BDCL) with the help ofcross-modal learning under bird's-eye view. We design three domaingeneralization settings based on three 3D datasets, and BEV-DG significantlyoutperforms state-of-the-art competitors with tremendous margins in allsettings.

BEV-DG: Cross-Modal Learning under Bird's-Eye View for Domain Generalization of 3D Semantic Segmentation