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

Real-world robotics applications demand object pose estimation methods thatwork reliably across a variety of scenarios. Modern learning-based approachesrequire large labeled datasets and tend to perform poorly outside the trainingdomain. Our first contribution is to develop a robust corrector module thatcorrects pose estimates using depth information, thus enabling existing methodsto better generalize to new test domains; the corrector operates on semantickeypoints (but is also applicable to other pose estimators) and is fullydifferentiable. Our second contribution is an ensemble self-training approachthat simultaneously trains multiple pose estimators in a self-supervisedmanner. Our ensemble self-training architecture uses the robust corrector torefine the output of each pose estimator; then, it evaluates the quality of theoutputs using observable correctness certificates; finally, it uses theobservably correct outputs for further training, without requiring externalsupervision. As an additional contribution, we propose small improvements to aregression-based keypoint detection architecture, to enhance its robustness tooutliers; these improvements include a robust pooling scheme and a robustcentroid computation. Experiments on the YCBV and TLESS datasets show theproposed ensemble self-training outperforms fully supervised baselines whilenot requiring 3D annotations on real data.

A Correct-and-Certify Approach to Self-Supervise Object Pose Estimators via Ensemble Self-Training