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 focuses on vision-based pose estimation for multiple rigid objectsplaced in clutter, especially in cases involving occlusions and objects restingon each other. Progress has been achieved recently in object recognition givenadvancements in deep learning. Nevertheless, such tools typically require alarge amount of training data and significant manual effort to label objects.This limits their applicability in robotics, where solutions must scale to alarge number of objects and variety of conditions. Moreover, the combinatorialnature of the scenes that could arise from the placement of multiple objects ishard to capture in the training dataset. Thus, the learned models might notproduce the desired level of precision required for tasks, such as roboticmanipulation. This work proposes an autonomous process for pose estimation thatspans from data generation to scene-level reasoning and self-learning. Inparticular, the proposed framework first generates a labeled dataset fortraining a Convolutional Neural Network (CNN) for object detection in clutter.These detections are used to guide a scene-level optimization process, whichconsiders the interactions between the different objects present in the clutterto output pose estimates of high precision. Furthermore, confident estimatesare used to label online real images from multiple views and re-train theprocess in a self-learning pipeline. Experimental results indicate that thisprocess is quickly able to identify in cluttered scenes physically-consistentobject poses that are more precise than the ones found by reasoning overindividual instances of objects. Furthermore, the quality of pose estimatesincreases over time given the self-learning process.

Physics-based Scene-level Reasoning for Object Pose Estimation in Clutter