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

Current popular backbones in computer vision, such as Vision Transformers(ViT) and ResNets are trained to perceive the world from 2D images. However, tomore effectively understand 3D structural priors in 2D backbones, we proposeMask3D to leverage existing large-scale RGB-D data in a self-supervisedpre-training to embed these 3D priors into 2D learned feature representations.In contrast to traditional 3D contrastive learning paradigms requiring 3Dreconstructions or multi-view correspondences, our approach is simple: weformulate a pre-text reconstruction task by masking RGB and depth patches inindividual RGB-D frames. We demonstrate the Mask3D is particularly effective inembedding 3D priors into the powerful 2D ViT backbone, enabling improvedrepresentation learning for various scene understanding tasks, such as semanticsegmentation, instance segmentation and object detection. Experiments show thatMask3D notably outperforms existing self-supervised 3D pre-training approacheson ScanNet, NYUv2, and Cityscapes image understanding tasks, with animprovement of +6.5% mIoU against the state-of-the-art Pri3D on ScanNet imagesemantic segmentation.

Mask3D: Pre-training 2D Vision Transformers by Learning Masked 3D Priors