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

Visual pre-training with large-scale real-world data has made great progressin recent years, showing great potential in robot learning with pixelobservations. However, the recipes of visual pre-training for robotmanipulation tasks are yet to be built. In this paper, we thoroughlyinvestigate the effects of visual pre-training strategies on robot manipulationtasks from three fundamental perspectives: pre-training datasets, modelarchitectures and training methods. Several significant experimental findingsare provided that are beneficial for robot learning. Further, we propose avisual pre-training scheme for robot manipulation termed Vi-PRoM, whichcombines self-supervised learning and supervised learning. Concretely, theformer employs contrastive learning to acquire underlying patterns fromlarge-scale unlabeled data, while the latter aims learning visual semantics andtemporal dynamics. Extensive experiments on robot manipulations in varioussimulation environments and the real robot demonstrate the superiority of theproposed scheme. Videos and more details can be found on\url{https://explore-pretrain-robot.github.io}.

Exploring Visual Pre-training for Robot Manipulation: Datasets, Models and Methods