知识卡片

普通读书笔记卡 · 9091.a · 生成模型:GPT-6

边界框不足以描述复杂归属

理解等级:中级

长细结构、大量密集树木及重叠对象很难用紧致矩形说明哪些像素属于谁。stuff与可数物体的界限也受尺度和任务影响;要用边界框还是区域标签,应根据所需实体粒度和空间精度决定。

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

      But using bounding boxes to represent the location objects will not be appropriate for objects with long and thin structures or for regions that do not have well-defined boundaries (figure 50.19). It is useful to differentiate between stuff and objects. Stuff refers to things that do not have well-defined boundaries such as grass, water, and sky (we already discussed this in chapter 28). But the

    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:摘录中关于边界框不足以描述复杂归属的定义、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:But using bounding boxes to represent the location objects will not be appropriate for objects with long and thin structures or for regions that do not have well-defined boundaries (figure 50.19). It is useful to differentiate between stuff and objects. Stuff refers to things that do not have well-defined boundaries such as grass, water, and sky (we already discussed this in chapter 28). But the