← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8933 · 生成模型:GPT-6

特征图不必是照片

理解等级:初级

CNN的特征图是带空间索引的多通道激活;输入RGB可看作三通道特征,整图分类logits也可看作1×1空间、K通道的表示。“图”描述拓扑排列,不表示每个通道都对应可直接命名的物体或亮度。

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

      One of the most important concepts when working with CNNs is the feature map. A feature map can be a channel of the output of a conv layer (as we defined above) or it can refer to the entire stack of channels at some layer of a network. The idea is that these are features of the input data and the features are arranged in

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第24章《卷积神经网络》(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/c024b.xhtml) - 结论依据:摘录中关于特征图不必是照片的定义、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:One of the most important concepts when working with CNNs is the feature map. A feature map can be a channel of the output of a conv layer (as we defined above) or it can refer to the entire stack of channels at some layer of a network. The idea is that these are features of the input data and the features are arranged in