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/

Zendy Plus

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

Zendy Tools

Zendy Open

As model and dataset sizes continue to scale in robot learning, the need tounderstand how the composition and properties of a dataset affect modelperformance becomes increasingly urgent to ensure cost-effective datacollection and model performance. In this work, we empirically investigate howphysics attributes (color, friction coefficient, shape) and scene backgroundcharacteristics, such as the complexity and dynamics of interactions withbackground objects, influence the performance of Video Transformers inpredicting planar pushing trajectories. We investigate three primary questions:How do physics attributes and background scene characteristics influence modelperformance? What kind of changes in attributes are most detrimental to modelgeneralization? What proportion of fine-tuning data is required to adapt modelsto novel scenarios? To facilitate this research, we presentCloudGripper-Push-1K, a large real-world vision-based robot pushing datasetcomprising 1278 hours and 460,000 videos of planar pushing interactions withobjects with different physics and background attributes. We also propose VideoOcclusion Transformer (VOT), a generic modular video-transformer-basedtrajectory prediction framework which features 3 choices of 2D-spatial encodersas the subject of our case study. The dataset and source code are available athttps://cloudgripper.org.

How Physics and Background Attributes Impact Video Transformers in Robotic Manipulation: A Case Study on Planar Pushing