← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9173.a · 生成模型:GPT-6

卷积输出尺寸要计算边界和步长

理解等级:中级

无膨胀的一维边长满足O=floor((I+2P−K)/S)+1,多维逐轴计算。28输入、K=3、P=0、S=1得到26,K=7得到22;保持尺寸需特定padding与stride条件,不能所有步长都只加半核宽就不变。

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

      As mentioned earlier, a convolution of kernel width 2r + 1 reduces the input by r on each side. If this is not desired (e.g., when multiple convolutional layers are used in series), padding can be used to surround the input image with up to r pixels, which results in the output image having the same dimension as the input image. Instead of moving the

    参考来源

    - 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml) - 结论依据:摘录中关于卷积输出尺寸要计算边界和步长的定义、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:As mentioned earlier, a convolution of kernel width 2r + 1 reduces the input by r on each side. If this is not desired (e.g., when multiple convolutional layers are used in series), padding can be used to surround the input image with up to r pixels, which results in the output image having the same dimension as the input image. Instead of moving the