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

Due to difficulties in acquiring ground truth depth of equirectangular (360)images, the quality and quantity of equirectangular depth data today isinsufficient to represent the various scenes in the world. Therefore, 360 depthestimation studies, which relied solely on supervised learning, are destined toproduce unsatisfactory results. Although self-supervised learning methodsfocusing on equirectangular images (EIs) are introduced, they often haveincorrect or non-unique solutions, causing unstable performance. In this paper,we propose 360 monocular depth estimation methods which improve on the areasthat limited previous studies. First, we introduce a self-supervised 360 depthlearning method that only utilizes gravity-aligned videos, which has thepotential to eliminate the needs for depth data during the training procedure.Second, we propose a joint learning scheme realized by combining supervised andself-supervised learning. The weakness of each learning is compensated, thusleading to more accurate depth estimation. Third, we propose a non-local fusionblock, which can further retain the global information encoded by visiontransformer when reconstructing the depths. With the proposed methods, wesuccessfully apply the transformer to 360 depth estimations, to the best of ourknowledge, which has not been tried before. On several benchmarks, our approachachieves significant improvements over previous works and establishes a stateof the art.

Improving 360 Monocular Depth Estimation via Non-local Dense Prediction Transformer and Joint Supervised and Self-supervised Learning