← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9045 · 生成模型:GPT-6

已知物体尺寸可引入尺度

理解等级:中级

熟悉对象的真实尺寸与像大小结合相机模型,可为绝对深度提供参照。人或家具只有统计上的典型尺寸时,尺度估计仍有个体与姿态误差;类别识别本身不提供精确真实长度。需要明确使用实测尺寸还是类别先验。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Depth can also be estimated by detecting familiar objects with known sizes. One example is people detection. Objects with known sizes can be used to estimate absolute depth while many of the cues that we discussed before can only give relative depth measurements. One example of using object sizes to reason about 3D geometry is [213].

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第43章《单图深度估计》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c043.xhtml) - 结论依据:摘录中关于已知物体尺寸可引入尺度的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Depth can also be estimated by detecting familiar objects with known sizes. One example is people detection. Objects with known sizes can be used to estimate absolute depth while many of the cues that we discussed before can only give relative depth measurements. One example of using object sizes to reason about 3D geometry is [213].