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

As transformer architectures and dataset sizes continue to scale, the need tounderstand the specific dataset factors affecting model performance becomesincreasingly urgent. This paper investigates how object physics attributes(color, friction coefficient, shape) and background characteristics (static,dynamic, background complexity) influence the performance of Video Transformersin trajectory prediction tasks under occlusion. Beyond mere occlusionchallenges, this study aims to investigate three questions: How do objectphysics attributes and background characteristics influence the modelperformance? What kinds of attributes are most influential to the modelgeneralization? Is there a data saturation point for large transformer modelperformance within a single task? To facilitate this research, we presentOccluManip, a real-world video-based robot pushing dataset comprising 460,000consistent recordings of objects with different physics and varyingbackgrounds. 1.4 TB and in total 1278 hours of high-quality videos of flexibletemporal length along with target object trajectories are collected,accommodating tasks with different temporal requirements. Additionally, wepropose Video Occlusion Transformer (VOT), a generic video-transformer-basednetwork achieving an average 96% accuracy across all 18 sub-datasets providedin OccluManip. OccluManip and VOT will be released at:https://github.com/ShutongJIN/OccluManip.git

Video Transformers under Occlusion: How Physics and Background Attributes Impact Large Models for Robotic Manipulation