← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9067 · 生成模型:GPT-6

正确对应外观也会变化

理解等级:中级

车标示例中最佳对应并非与输入块逐像素相同,原因包括亚像素运动及物体接近后的尺度变化。非零最佳代价不能单独证明错误,对应质量需结合候选间差距、几何与运动连续性判断。

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

      In the example shown in figure 46.4, the input patch is centered on the car logo. The matching cost displays the distance using a reversed grayscale map, with smaller values appearing as brighter spots. In this case, the search identifies a unique match. However, it is worth noting that the best matching patch, although correctly detecting the same logo, is not identical to the input

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第46章《运动估计》(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/c046b.xhtml) - 结论依据:摘录中关于正确对应外观也会变化的定义、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:In the example shown in figure 46.4, the input patch is centered on the car logo. The matching cost displays the distance using a reversed grayscale map, with smaller values appearing as brighter spots. In this case, the search identifies a unique match. However, it is worth noting that the best matching patch, although correctly detecting the same logo, is not identical to the input