第七课 TensorFlow实现卷积神经网络

; TensorFlow实现卷积神经网络

一 本节课程介绍

1.1 知识点

1、卷积神经网络介绍;
2、TensorFlow实践CNN网络;

二 课程内容

2.1 卷积神经网络基本介绍

卷积神经网络是一种利用卷积结构构造的神经网络模型,具有局部感知、权值分担、参数归约和多层次结构等特点。

[En]

Convolution neural network is a kind of neural network model constructed by convolution structure, which is characterized by local perception, weight sharing, pooling reduction of parameters and multi-level structure.

其基本结构包括输入层、卷积层、下采样层和全连接输出层。每一层都用卷积检查图像进行卷积,计算出的矩阵称为特征图,称为原始图像映射区域的感受野。一般而言,第一卷积层的接受场大小等于卷积核大小,而后续卷积层的接受场大小与前一卷积核大小和步长有关。让我们从卷积核和步长的基本概念开始。

[En]

Its basic structure includes input layer, convolution layer, downsampling layer and fully connected output layer. Each layer is convoluted by the convolution check image, and the calculated matrix is called the feature graph, which is called the receptive field in the area mapped by the original image. Generally speaking, the receptive field size of the first convolution layer is equal to the convolution core size, while the receptive field size of the subsequent convolution layer is related to the previous convolution core size and step size. Let’s start with the basic concepts of convolution kernel and step size.

2.1.1 卷积核和步长

卷积核包括卷积核大小、输入通道数和输出通道数。例如,对具有32个通道和输入的64个卷积核进行卷积,以获得64个卷积结果。卷积是通过将元素逐个相乘和求和来计算的。每个卷积运动的长度就是它的卷积步长。

[En]

The convolution kernel includes convolution kernel size, the number of input channels and the number of output channels. For example, 64 convolution kernels with 32 channels and inputs are convoluted to get 64 convolution results. The convolution is calculated by multiplying and summing elements one by one. The length of each convolution movement is its convolution step.

第七课 TensorFlow实现卷积神经网络

Original: https://blog.csdn.net/qq_42279468/article/details/122696083
Author: 大龙剑神
Title: 第七课 TensorFlow实现卷积神经网络

原创文章受到原创版权保护。转载请注明出处:https://www.johngo689.com/511494/

转载文章受原作者版权保护。转载请注明原作者出处!

(0)

大家都在看

亲爱的 Coder【最近整理,可免费获取】👉 最新必读书单  | 👏 面试题下载  | 🌎 免费的AI知识星球