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

Effectively utilizing the vast amounts of ego-centric navigation data that isfreely available on the internet can advance generalized intelligent systems,i.e., to robustly scale across perspectives, platforms, environmentalconditions, scenarios, and geographical locations. However, it is difficult todirectly leverage such large amounts of unlabeled and highly diverse data forcomplex 3D reasoning and planning tasks. Consequently, researchers haveprimarily focused on its use for various auxiliary pixel- and image-levelcomputer vision tasks that do not consider an ultimate navigational objective.In this work, we introduce SelfD, a framework for learning scalable driving byutilizing large amounts of online monocular images. Our key idea is to leverageiterative semi-supervised training when learning imitative agents fromunlabeled data. To handle unconstrained viewpoints, scenes, and cameraparameters, we train an image-based model that directly learns to plan in theBird's Eye View (BEV) space. Next, we use unlabeled data to augment thedecision-making knowledge and robustness of an initially trained model viaself-training. In particular, we propose a pseudo-labeling step which enablesmaking full use of highly diverse demonstration data through "hypothetical"planning-based data augmentation. We employ a large dataset of publiclyavailable YouTube videos to train SelfD and comprehensively analyze itsgeneralization benefits across challenging navigation scenarios. Withoutrequiring any additional data collection or annotation efforts, SelfDdemonstrates consistent improvements (by up to 24%) in driving performanceevaluation on nuScenes, Argoverse, Waymo, and CARLA.

SelfD: Self-Learning Large-Scale Driving Policies From the Web