知识卡片

普通读书笔记卡 · 9083.b · 生成模型:GPT-6

组件组合提供可复用表示

理解等级:中级

按组件识别理论用少量几何部件及其关系描述多类对象,非偶然属性有助于识别部件。它是有启发性的理论与建模方案,不是已证明覆盖所有人类物体识别的唯一机制;纹理、草、云等也未必适合少数组件表达。

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

      Biederman postulated that objects are represented by decomposing the object into a small set of geometric components (e.g., blocks, cylinders, cones, and so on.) that he called Geons. He derived 36 different Geons that we can discriminate visually. Geons we defined by non-accidental attributes such as symmetry, colinearity, cotermination, parallelism, and curvature. An object is defined by a particular arrangement of a small set of

    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:摘录中关于组件组合提供可复用表示的定义、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:Biederman postulated that objects are represented by decomposing the object into a small set of geometric components (e.g., blocks, cylinders, cones, and so on.) that he called Geons. He derived 36 different Geons that we can discriminate visually. Geons we defined by non-accidental attributes such as symmetry, colinearity, cotermination, parallelism, and curvature. An object is defined by a particular arrangement of a small set of