深度学习论文: Depth Anything V2
Depth Anything V2
PDF: https://arxiv.org/pdf/2406.09414v1
代码:https://depth-anything-v2.github.io/
PyTorch代码: https://github.com/shanglianlm0525/CvPytorch
PyTorch代码: https://github.com/shanglianlm0525/PyTorch-Networks
1 概述
Depth Anything V2通过三项关键实践实现了更精细、更稳健的单目深度估计:首先,采用合成图像替代真实标记图像,以克服标签噪声和细节缺失;其次,扩大教师模型的容量,提升性能;最后,利用大规模伪标签真实图像训练学生模型。与基于Stable Diffusi