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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The advent of autonomous driving and advanced driver assistance systemsnecessitates continuous developments in computer vision for 3D sceneunderstanding. Self-supervised monocular depth estimation, a method forpixel-wise distance estimation of objects from a single camera without the useof ground truth labels, is an important task in 3D scene understanding.However, existing methods for this task are limited to convolutional neuralnetwork (CNN) architectures. In contrast with CNNs that use localized linearoperations and lose feature resolution across the layers, vision transformersprocess at constant resolution with a global receptive field at every stage.While recent works have compared transformers against their CNN counterpartsfor tasks such as image classification, no study exists that investigates theimpact of using transformers for self-supervised monocular depth estimation.Here, we first demonstrate how to adapt vision transformers for self-supervisedmonocular depth estimation. Thereafter, we compare the transformer andCNN-based architectures for their performance on KITTI depth predictionbenchmarks, as well as their robustness to natural corruptions and adversarialattacks, including when the camera intrinsics are unknown. Our studydemonstrates how transformer-based architecture, though lower in run-timeefficiency, achieves comparable performance while being more robust andgeneralizable.

Transformers in Self-Supervised Monocular Depth Estimation with Unknown Camera Intrinsics