← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9088.b · 生成模型:GPT-6

级联计算热点可落在难负例

理解等级:中级

热力图中耗费最多计算的区域可能是最终错误候选,因为它们难以在早期拒绝。计算分配不是对象存在性的独立证据;应检查难负例在哪一层保留,以及它们如何影响时延与误报。

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

      Figure 50.14: (left) Input image with the output of a face detector from 1999. (right) The heat map reports the computational cost at each location. Most of the computation was allocated into the wrong detections in the top left corner. Figure from [132].

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第50章《目标识别》(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/c050b.xhtml) - 结论依据:摘录中关于级联计算热点可落在难负例的定义、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:Figure 50.14: (left) Input image with the output of a face detector from 1999. (right) The heat map reports the computational cost at each location. Most of the computation was allocated into the wrong detections in the top left corner. Figure from [132].