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

In bin-picking scenarios, multiple instances of an object of interest arestacked in a pile randomly, and hence, the instances are inherently subjectedto the challenges: severe occlusion, clutter, and similar-looking distractors.Most existing methods are, however, for single isolated object instances, whilesome recent methods tackle crowd scenarios as post-refinement which accountsmultiple object relations. In this paper, we address recovering 6D poses ofmultiple instances in bin-picking scenarios in depth modality by multi-tasklearning in deep neural networks. Our architecture jointly learns multiplesub-tasks: 2D detection, depth, and 3D pose estimation of individual objects;and joint registration of multiple objects. For training data generation, depthimages of physically plausible object pose configurations are generated by a 3Dobject model in a physics simulation, which yields diverse occlusion patternsto learn. We adopt a state-of-the-art object detector, and 2D offsets arefurther estimated via a network to refine misaligned 2D detections. The depthand 3D pose estimator is designed to generate multiple hypotheses perdetection. This allows the joint registration network to learn occlusionpatterns and remove physically implausible pose hypotheses. We apply ourarchitecture on both synthetic (our own and Sileane dataset) and real (a publicBin-Picking dataset) data, showing that it significantly outperformsstate-of-the-art methods by 15-31% in average precision.

Multi-Task Deep Networks for Depth-Based 6D Object Pose and Joint Registration in Crowd Scenarios