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 work proposes a process for efficiently searching over combinations ofindividual object 6D pose hypotheses in cluttered scenes, especially in casesinvolving occlusions and objects resting on each other. The initial set ofcandidate object poses is generated from state-of-the-art object detection andglobal point cloud registration techniques. The best-scored pose per object byusing these techniques may not be accurate due to overlaps and occlusions.Nevertheless, experimental indications provided in this work show that objectposes with lower ranks may be closer to the real poses than ones with highranks according to registration techniques. This motivates a globaloptimization process for improving these poses by taking into accountscene-level physical interactions between objects. It also implies that theCartesian product of candidate poses for interacting objects must be searchedso as to identify the best scene-level hypothesis. To perform the searchefficiently, the candidate poses for each object are clustered so as to reducetheir number but still keep a sufficient diversity. Then, searching over thecombinations of candidate object poses is performed through a Monte Carlo TreeSearch (MCTS) process that uses the similarity between the observed depth imageof the scene and a rendering of the scene given the hypothesized pose as ascore that guides the search procedure. MCTS handles in a principled way thetradeoff between fine-tuning the most promising poses and exploring new ones,by using the Upper Confidence Bound (UCB) technique. Experimental resultsindicate that this process is able to quickly identify in cluttered scenesphysically-consistent object poses that are significantly closer to groundtruth compared to poses found by point cloud registration methods.

Improving 6D Pose Estimation of Objects in Clutter via Physics-aware Monte Carlo Tree Search