CNN中感受野的计算 | 小王同学

https://a171232886.github.io/2021/11/23/CNN%E4%B8%AD%E6%84%9F%E5%8F%97%E9%87%8E%E7%9A%84%E8%AE%A1%E7%AE%97/

概念
感受野 (receptive field, RF)是卷积神经网络每一层输出的特征图上的像素点在原始图像上映射的区域大小。
在网上看到了一张很好的图，可以表示整个关系。



图片来源
注意：计算感受野大小时，忽略了图像边缘的影响，即不考虑padding的大小。

The issue links to a blog post about calculating receptive fields in CNNs. The repository appears to be a personal blog using Hexo. The issue does not specify what change is needed. To start, examine the linked post and the repository's content structure to see if the post is already present or needs to be added or updated. Check the source files in the Hexo blog's post directory. Determine if the task is to add this post, fix a link, or something else.

Scritto dal modello di indicizzazione a partire dal testo della issue.