SN Applied Sciences (Nov 2021)

Detail-preserving depth estimation from a single image based on modified fully convolutional residual network and gradient network

  • Huihui Xu,
  • Nan Liu

DOI
https://doi.org/10.1007/s42452-021-04882-0
Journal volume & issue
Vol. 3, no. 12
pp. 1 – 15

Abstract

Read online

Article Highlights We changed the classic network and loss function to obtain the global 3D depth information of the scene. A depth gradient acquisition scheme is designed to generate the local details of the scene. We can obtain a plausible depth map with better depth details through our developed depth and depth gradients fusion strategy.

Keywords