<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Tensorflow on Mingy'Log</title><link>https://humyue.github.io/categories/tensorflow/</link><description>Recent content in Tensorflow on Mingy'Log</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Thu, 08 Mar 2018 10:22:54 +0800</lastBuildDate><atom:link href="https://humyue.github.io/categories/tensorflow/index.xml" rel="self" type="application/rss+xml"/><item><title>卷积神经网络中两种Padding方式的介绍</title><link>https://humyue.github.io/posts/two-padding/</link><pubDate>Thu, 08 Mar 2018 10:22:54 +0800</pubDate><guid>https://humyue.github.io/posts/two-padding/</guid><description>&lt;p&gt;当我们计算卷积池化后的图像尺寸时，需要注意“VALID”和“SAME”两种Padding对应的输出尺寸计算方式不一样，下面就来说说具体的计算方式&lt;/p&gt;
&lt;h3 id="1-当padding为valid"&gt;1. 当Padding为“VALID”&lt;/h3&gt;
&lt;p&gt;这种情况下，当你经过最后一步步长做卷积或池化运算，超出图像边界不会补‘0’，输出的形状为：⌈(W-F+1)/S⌉&lt;/p&gt;</description></item></channel></rss>