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

Pose estimation of 3D objects in monocular images is a fundamental andlong-standing problem in computer vision. Existing deep learning approaches for6D pose estimation typically rely on the assumption of availability of 3Dobject models and 6D pose annotations. However, precise annotation of 6D posesin real data is intricate, time-consuming and not scalable, while syntheticdata scales well but lacks realism. To avoid these problems, we present aweakly-supervised reconstruction-based pipeline, named NeRF-Pose, which needsonly 2D object segmentation and known relative camera poses during training.Following the first-reconstruct-then-regress idea, we first reconstruct theobjects from multiple views in the form of an implicit neural representation.Then, we train a pose regression network to predict pixel-wise 2D-3Dcorrespondences between images and the reconstructed model. At inference, theapproach only needs a single image as input. A NeRF-enabled PnP+RANSACalgorithm is used to estimate stable and accurate pose from the predictedcorrespondences. Experiments on LineMod and LineMod-Occlusion show that theproposed method has state-of-the-art accuracy in comparison to the best 6D poseestimation methods in spite of being trained only with weak labels. Besides, weextend the Homebrewed DB dataset with more real training images to support theweakly supervised task and achieve compelling results on this dataset. Theextended dataset and code will be released soon.

NeRF-Pose: A First-Reconstruct-Then-Regress Approach for Weakly-supervised 6D Object Pose Estimation