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

Depth and ego-motion estimations are essential for the localization andnavigation of autonomous robots and autonomous driving. Recent studies make itpossible to learn the per-pixel depth and ego-motion from the unlabeledmonocular video. A novel unsupervised training framework is proposed with 3Dhierarchical refinement and augmentation using explicit 3D geometry. In thisframework, the depth and pose estimations are hierarchically and mutuallycoupled to refine the estimated pose layer by layer. The intermediate viewimage is proposed and synthesized by warping the pixels in an image with theestimated depth and coarse pose. Then, the residual pose transformation can beestimated from the new view image and the image of the adjacent frame to refinethe coarse pose. The iterative refinement is implemented in a differentiablemanner in this paper, making the whole framework optimized uniformly.Meanwhile, a new image augmentation method is proposed for the pose estimationby synthesizing a new view image, which creatively augments the pose in 3Dspace but gets a new augmented 2D image. The experiments on KITTI demonstratethat our depth estimation achieves state-of-the-art performance and evensurpasses recent approaches that utilize other auxiliary tasks. Our visualodometry outperforms all recent unsupervised monocular learning-based methodsand achieves competitive performance to the geometry-based method, ORB-SLAM2with back-end optimization.

3D Hierarchical Refinement and Augmentation for Unsupervised Learning of Depth and Pose from Monocular Video