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

This paper introduces Action Image, a new grasp proposal representation thatallows learning an end-to-end deep-grasping policy. Our model achieves $84\%$grasp success on $172$ real world objects while being trained only insimulation on $48$ objects with just naive domain randomization. Similar tocomputer vision problems, such as object detection, Action Image builds on theidea that object features are invariant to translation in image space.Therefore, grasp quality is invariant when evaluating the object-gripperrelationship; a successful grasp for an object depends on its local context,but is independent of the surrounding environment. Action Image represents agrasp proposal as an image and uses a deep convolutional network to infer graspquality. We show that by using an Action Image representation, trained networksare able to extract local, salient features of grasping tasks that generalizeacross different objects and environments. We show that this representationworks on a variety of inputs, including color images (RGB), depth images (D),and combined color-depth (RGB-D). Our experimental results demonstrate thatnetworks utilizing an Action Image representation exhibit strong domaintransfer between training on simulated data and inference on real-world sensorstreams. Finally, our experiments show that a network trained with Action Imageimproves grasp success ($84\%$ vs. $53\%$) over a baseline model with the samestructure, but using actions encoded as vectors.

Action Image Representation: Learning Scalable Deep Grasping Policies with Zero Real World Data