利用结构与语义的无监督单目深度与自运动学习
计算机视觉与模式识别
2019-06-14 v1 机器人学
摘要
我们提出了一种利用结构与语义进行无监督单目深度与自运动学习的方法。更具体地，我们对单个物体的运动进行建模，并将其 3D 运动向量与深度和自运动联合学习。我们获得了更精确的结果，尤其是对于先前方法未解决的具有挑战性的动态场景。此为 Casser 等人 [AAAI'19] 的扩展版本。代码与模型已开源：https://sites.google.com/corp/view/struct2depth。
引用
@article{arxiv.1906.05717,
title = {Unsupervised Monocular Depth and Ego-motion Learning with Structure and Semantics},
author = {Vincent Casser and Soeren Pirk and Reza Mahjourian and Anelia Angelova},
journal= {arXiv preprint arXiv:1906.05717},
year = {2019}
}
备注
CVPR Workshop on Visual Odometry & Computer Vision Applications Based on Location Clues (VOCVALC), 2019. This is an extension of arXiv:1811.06152: Depth Prediction Without the Sensors: Leveraging Structure for Unsupervised Learning from Monocular Videos. Thirty-Third AAAI Conference on Artificial Intelligence (AAAI'19)