Deep Residual Learning for Image Recognition - Githubissues

JasonWayne / deep-learning-essay

0 stars 0 forks source link

Deep Residual Learning for Image Recognition #24

Open JasonWayne opened 4 years ago

JasonWayne commented 4 years ago

ResNet50

https://www.semanticscholar.org/paper/Deep-Residual-Learning-for-Image-Recognition-He-Zhang/29c808b346526fbb6027e67942b62a40a549f019

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7780459&tag=1

JasonWayne commented 4 years ago

头部结构（以ResNet50为例）：

标准的ResNet50，头部是一个stride=2，7x7，channel=64的大卷积，跟着一个stride=2的MaxPooling结构。

NAS中的ResNet50，头部一般是两个3x3，stride=2，channel分别为32，64的两层卷积。

JasonWayne commented 4 years ago

FeatureMap大小的变化：头部除了四，后面除了第一个stage以外，每个stage的第一个block会除以2，