📡 Embodied AI Radar具身智能雷达

Topics主题 › Perception, SLAM & 3D Mapping感知、SLAM 与三维建图

Benchmark基准 2026 new新 2026-10-21

Which Depth Estimator for Jetson AGX Orin? A 2026 Benchmark of Monocular and Stereo Methods

Measured numbers for robot builders: nine monocular and stereo depth families (Depth Anything V2/3, Depth Pro, MoGe-2, Fast-FoundationStereo, ...) run on a Jetson AGX Orin, scored for metric accuracy and speed.给机器人开发者的实测数据:在 Jetson AGX Orin 上用 TensorRT 运行九类单目与双目深度估计模型(Depth Anything V2/3、Depth Pro、MoGe-2、Fast-FoundationStereo 等),并评测度量精度与速度。

BibTeX
@misc{jiang2026which,
  title = {{Which Depth Estimator for Jetson AGX Orin? A 2026 Benchmark of Monocular and Stereo Methods}},
  author = {Leaf Jiang},
  year = {2026},
  eprint = {2610.27199},
  archivePrefix = {arXiv},
  primaryClass = {cs.RO},
  url = {https://arxiv.org/abs/2610.27199},
}

Authors (from arXiv): 作者(来自 arXiv):Leaf Jiang

More in Perception, SLAM & 3D Mapping更多感知、SLAM 与三维建图条目

All Perception, SLAM & 3D Mapping entries →查看全部感知、SLAM 与三维建图条目 →