GeoCueDepth: Exploiting Geometric Structure Cues to Estimate Depth from a Single Image

Yiming Zeng1, Yu Hu2, Shice Liu3, Qiankun Tang4, Jing Ye5, Xiaowei Li5

  • 1Institude of Computing Technology, Chinese Academy of Sciences
  • 2Institute of Computing Technology Chinese Academy of Sciences
  • 3Institute Of Computing Technology Chinese Academy Of Sciences
  • 4The Institute of Computing Technology of the Chinese Academy of
  • 5Institute of Computing Technology, Chinese Academy of Sciences

Details

11:00 - 11:15 | Mon 25 Sep | Room 109 | MoAT1.3

Session: Deep Learning in Robotics and Automation I

Abstract

Depth estimation from a single image is very challenging due to the inherent ambiguity of mapping a color image to a depth map. Previous work tackles this problem by exploiting various levels of features with multi-scale deep convolutional neural networks. However, most of the local geometric structure related monocular depth cues are lost when being propagated through convolutional neural network. Moreover, the error of depth cues related to local geometric structures is not considered in the loss function. In this work, we propose the GeoCueDepth convolutional neural network to exploit local geometric structure cues and propose a training loss that takes the geometric error into consideration, which significantly improve the performance of depth prediction in both accuracy and sharpness. Experiments show that the proposed method achieves 0.122 average relative error and 0.078 square relative error on the NYU Depth v2 data set, which outperforms state-of-the-art monocular depth estimation approaches.