知识卡片

普通读书笔记卡 · 8973 · 生成模型:GPT-6

聚类几何由特征尺度决定

理解等级:中级

k均值使用特征空间里的欧氏平方距离,特征缩放或编码器改变会改变最近中心及分组结果。因此视觉分组是否发现轮廓、表面或对象,依赖所用表示与相容目标,不能把像素直接聚成若干组就等同发现语义对象。

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 vision, the problem of clustering is related to the idea of perceptual grouping, which we cover in detail in chapter 31. We humans see the world as organized into different levels of perceptual structure: contours and surfaces, objects and events. These structures are groupings, or clusters, of related visual elements: a contour is a group of points that form a line, an object is

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第30章《表示学习》(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/c030b.xhtml) - 结论依据:摘录中关于聚类几何由特征尺度决定的定义、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 vision, the problem of clustering is related to the idea of perceptual grouping, which we cover in detail in chapter 31. We humans see the world as organized into different levels of perceptual structure: contours and surfaces, objects and events. These structures are groupings, or clusters, of related visual elements: a contour is a group of points that form a line, an object is