← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8817 · 生成模型:GPT-6

视觉是从光推断世界属性

理解等级:初级

视觉接收的是从不同方向到达传感器的光,目标却是推断表面、物体、材质与可行动信息。输入强度与期望输出不是同一种量,而且投影、反射和遮挡会丢失信息。因此识别场景不是直接读出像素已有的标签,需要假设、知识与任务约束。

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

      The goal of vision is not to measure light intensities, but to extract scene properties relevant for the observer. Figure 1.2 shows different images that illustrate that the human visual system is trying to recover the scenes that are the cause of those images.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第01章《视觉的挑战》(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/c001.xhtml) - 结论依据:摘录中关于视觉是从光推断世界属性的定义、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:The goal of vision is not to measure light intensities, but to extract scene properties relevant for the observer. Figure 1.2 shows different images that illustrate that the human visual system is trying to recover the scenes that are the cause of those images.