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

Learning a generalizable object manipulation policy is vital for an embodiedagent to work in complex real-world scenes. Parts, as the shared components indifferent object categories, have the potential to increase the generalizationability of the manipulation policy and achieve cross-category objectmanipulation. In this work, we build the first large-scale, part-basedcross-category object manipulation benchmark, PartManip, which is composed of11 object categories, 494 objects, and 1432 tasks in 6 task classes. Comparedto previous work, our benchmark is also more diverse and realistic, i.e.,having more objects and using sparse-view point cloud as input without oracleinformation like part segmentation. To tackle the difficulties of vision-basedpolicy learning, we first train a state-based expert with our proposedpart-based canonicalization and part-aware rewards, and then distill theknowledge to a vision-based student. We also find an expressive backbone isessential to overcome the large diversity of different objects. Forcross-category generalization, we introduce domain adversarial learning fordomain-invariant feature extraction. Extensive experiments in simulation showthat our learned policy can outperform other methods by a large margin,especially on unseen object categories. We also demonstrate our method cansuccessfully manipulate novel objects in the real world.

PartManip: Learning Cross-Category Generalizable Part Manipulation Policy from Point Cloud Observations