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

Learning-based monocular depth estimation leverages geometric priors presentin the training data to enable metric depth perception from a single image, atraditionally ill-posed problem. However, these priors are often specific to aparticular domain, leading to limited generalization performance on unseendata. Apart from the well studied environmental domain gap, monocular depthestimation is also sensitive to the domain gap induced by varying cameraparameters, an aspect that is often overlooked in current state-of-the-artapproaches. This issue is particularly evident in autonomous driving scenarios,where datasets are typically collected with a single vehicle-camera setup,leading to a bias in the training data due to a fixed perspective geometry. Inthis paper, we challenge this trend and introduce GenDepth, a novel modelcapable of performing metric depth estimation for arbitrary vehicle-camerasetups. To address the lack of data with sufficiently diverse cameraparameters, we first create a bespoke synthetic dataset collected withdifferent vehicle-camera systems. Then, we design GenDepth to simultaneouslyoptimize two objectives: (i) equivariance to the camera parameter variations onsynthetic data, (ii) transferring the learned equivariance to real-worldenvironmental features using a single real-world dataset with a fixedvehicle-camera system. To achieve this, we propose a novel embedding of cameraparameters as the ground plane depth and present a novel architecture thatintegrates these embeddings with adversarial domain alignment. We validateGenDepth on several autonomous driving datasets, demonstrating itsstate-of-the-art generalization capability for different vehicle-camerasystems.

GenDepth: Generalizing Monocular Depth Estimation for Arbitrary Camera Parameters via Ground Plane Embedding