BriefGPT.xyz
Apr, 2019
基于递归神经网络的单目视觉里程计和深度（非）监督学习
Recurrent Neural Network for (Un-)supervised Learning of Monocular VideoVisual Odometry and Depth
HTML
PDF
Rui Wang, Stephen M. Pizer, Jan-Michael Frahm
TL;DR
本研究提出了一种利用循环神经网络和多视图图像重投影以及前向-后向流一致性损失来训练的学习型多视图稠密深度地图和里程计估计方法，可用于视频的深度和视觉里程计估计，产生优于现有技术的单视图和多视图深度估计结果。
Abstract
deep learning
-based, single-view
depth estimation
methods have recently shown highly promising results. However, such methods ignore one of the most important features for determining depth in the human vision sy
→