← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9036.g · 生成模型:GPT-6

网络回归不改变单应假设

理解等级:中级

可以由网络根据双图回归H,替代或辅助对应点拟合,但输出仍受单应可表达范围限制。非平面视差、退化视图与域变化不会因换为学习器自动消失。评估应包含几何误差、有效范围与是否存在合适的单应模型。

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

      Homographies are an important class of geometric transforms between images and they have lots of applications. Homographies can be used to take measures on a plane given a reference measure. Although we have focused on computing homographies between images using correspondences, one can compute homographies in other ways. For instance, to produce a bird's-eye view of a plane one can compute the needed homography using

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第41章《单应性》(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/c041.xhtml) - 结论依据:摘录中关于网络回归不改变单应假设的定义、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:Homographies are an important class of geometric transforms between images and they have lots of applications. Homographies can be used to take measures on a plane given a reference measure. Although we have focused on computing homographies between images using correspondences, one can compute homographies in other ways. For instance, to produce a bird's-eye view of a plane one can compute the needed homography using