研究论文 · 2026

FlexDepth：面向鲁棒驾驶感知的灵活尺度驱动自监督单目深度估计

Towards Robust Driving Perception: A Flexible Scale-Driven Family for Self-Supervised Monocular Depth Estimation

Zhaowen Zhu, Li Zhang, Yujie Chen, Tian Zhang, Yingjie Wang, Mingxia Zhan · Springer Nature Switzerland · English

资料概述

FlexDepth 分离静态与动态训练，并按模型尺度调整解码器组件，通过驾驶基准和移动端评估研究复杂道路环境中的单目深度估计效率、置信处理与泛化。

以下为 arXiv 上作者提供的信息。

Computer Vision - ECCV 2026, Lecture Notes in Computer Science, vol. 17003, Springer, Cham, 2026

Accepted by ECCV2026. Code is available at https://github.com/startnew/flexdepth

版本: arXiv v3 — manuscript revision