Publication detail

Vision UFormer: Long-Range Monocular Absolute Depth Estimation

POLÁŠEK, T. ČADÍK, M. KELLER, Y. BENEŠ, B.

Original Title

Vision UFormer: Long-Range Monocular Absolute Depth Estimation

Type

journal article in Web of Science

Language

English

Original Abstract

We introduce Vision UFormer (ViUT), a novel deep neural long-range monocular depth estimator. The input is an RGB image, and the output is an image that stores the absolute distance of the object in the scene as its per-pixel values. ViUT consists of a Transformer encoder and a ResNet decoder combined with UNet style of skip connections. It is trained on 1M images across ten datasets in a staged regime that starts with easier-to-predict data such as indoor photographs and continues to more complex long-range outdoor scenes. We show that ViUT provides comparable results for normalized relative distances and short-range classical datasets such as NYUv2 and KITTI. We further show that it successfully estimates of absolute long-range depth in meters. We validate ViUT on a wide variety of long-range scenes showing its high estimation capabilities with a relative improvement of up to 23%. Absolute depth estimation finds application in many areas, and we show its usability in image composition, range annotation, defocus, and scene reconstruction.

Keywords

Absolute Depth Estimation, Monocular Depth Prediction, Long Range Distance, Transformer, UNet, Staged Training

Authors

POLÁŠEK, T.; ČADÍK, M.; KELLER, Y.; BENEŠ, B.

Released

26. 2. 2023

Publisher

Elsevier

Location

Oxford

ISBN

0097-8493

Periodical

COMPUTERS & GRAPHICS-UK

Year of study

111

Number

4

State

United Kingdom of Great Britain and Northern Ireland

Pages from

180

Pages to

189

Pages count

10

URL

https://www.sciencedirect.com/science/article/pii/S0097849323000262

BibTex

@article{BUT185048,
  author="Tomáš {Polášek} and Martin {Čadík} and Yosi {Keller} and Bedřich {Beneš}",
  title="Vision UFormer: Long-Range Monocular Absolute Depth Estimation",
  journal="COMPUTERS & GRAPHICS-UK",
  year="2023",
  volume="111",
  number="4",
  pages="180--189",
  doi="10.1016/j.cag.2023.02.003",
  issn="0097-8493",
  url="https://www.sciencedirect.com/science/article/pii/S0097849323000262"
}