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

This paper considers the problem of single image depth estimation. Theemployment of convolutional neural networks (CNNs) has recently brought aboutsignificant advancements in the research of this problem. However, mostexisting methods suffer from loss of spatial resolution in the estimated depthmaps; a typical symptom is distorted and blurry reconstruction of objectboundaries. In this paper, toward more accurate estimation with a focus ondepth maps with higher spatial resolution, we propose two improvements toexisting approaches. One is about the strategy of fusing features extracted atdifferent scales, for which we propose an improved network architectureconsisting of four modules: an encoder, decoder, multi-scale feature fusionmodule, and refinement module. The other is about loss functions for measuringinference errors used in training. We show that three loss terms, which measureerrors in depth, gradients and surface normals, respectively, contribute toimprovement of accuracy in an complementary fashion. Experimental results showthat these two improvements enable to attain higher accuracy than the currentstate-of-the-arts, which is given by finer resolution reconstruction, forexample, with small objects and object boundaries.

Revisiting Single Image Depth Estimation: Toward Higher Resolution Maps with Accurate Object Boundaries