CNN

注意, s2到c3采用了16个 6X5X5 的kernel,生成了16个feature map

Alexnet

卷积

一维卷积:

二维卷积: 对kernel有个反转操作,所以有的会生成对称的kernel,这样直接运算了。

参考佳文

四大经典CNN网络技术原理arrow-up-right

CMU研究者探索新卷积方法:在实验中可媲美基准CNN(附实验代码arrow-up-right

https://arxiv.org/abs/1512.07108arrow-up-right Recent Advances in Convolutional Neural Networks

Convolutional Neural Networks (CNNs): An Illustrated Explanationarrow-up-right Convolutional Neural Networks (LeNet)arrow-up-right 卷积神经网络(CNN)arrow-up-right Deep learningarrow-up-right Convolution Neural Network (CNN) 原理与实现arrow-up-right Deep Learning论文笔记之(四)CNN卷积神经网络推导和实现arrow-up-right

变形卷积核、可分离卷积?卷积神经网络中十大拍案叫绝的操作arrow-up-right Deep Learning for Computer Vision – Introduction to Convolution Neural Networksarrow-up-right

PCANet训练过程arrow-up-right 【目标检测】Fast RCNN算法详解arrow-up-right

解读|Facebook 何凯明发大招:Mask R-CNN 狙击目标实例分割arrow-up-right

Focal Loss for Dense Object Detectionarrow-up-right

理解卷积: 图像卷积与滤波的一些知识点arrow-up-right 此文有用的信息量巨大 Lode's Computer Graphics Tutorial - Image Filteringarrow-up-right 卷积arrow-up-right Understanding Convolutionsarrow-up-right 我对卷积的理解arrow-up-right 这个挺不错的

[CV] 通俗理解『卷积』——从傅里叶变换到滤波器arrow-up-right Conv Nets: A Modular Perspectivearrow-up-right

CNN 中, 1X1卷积核到底有什么作用呢?arrow-up-right 多通道(比如RGB三通道)卷积过程arrow-up-right

Inception in CNNarrow-up-right

Deep Residual Network 深度残差网络arrow-up-right 秒懂!何凯明的深度残差网络PPT是这样的arrow-up-right

http://geek.csdn.net/news/detail/129128arrow-up-right

https://zhuanlan.zhihu.com/p/24774302arrow-up-right SPPNet-引入空间金字塔池化改进RCNN

How do these "neural network style transfer" tools work?arrow-up-right

UC伯克利大学AI实验室用一张单色图像生成高质量3D几何结构arrow-up-right

Last updated

Was this helpful?