14 深度学习-卷积
1.简述人工智能、机器学习和深度学习三者的联系与区别。
AI是设备或机器的总体描述,它以智能的方式运行,机器学习作为其子集,深度学习作为机器学习的子集。换句话说,机器学习和深度学习被归类为AI,但AI不一定是机器学习或深度学习。
2. 全连接神经网络与卷积神经网络的联系与区别。
DNN是指深度神经网络,它是一个很广的概念,某种意义上CNN、RNN、GAN等都属于其范畴之内。DNN与CNN(卷积神经网络)的区别是DNN特指全连接的神经元结构,并不包含卷积单元或是时间上的关联。
3.理解卷积计算。
以digit0为例,进行手工演算。
from sklearn.datasets import load_digits #小数据集8*8
digits = load_digits()
0 | 0 | 5 | 13 | 9 | 1 | 0 | 0 |
0 | 0 | 13 | 15 | 10 | 15 | 5 | 0 |
0 | 3 | 15 | 2 | 0 | 11 | 8 | 0 |
0 | 4 | 12 | 0 | 0 | 8 | 8 | 0 |
0 | 5 | 8 | 0 | 0 | 9 | 8 | 0 |
0 | 4 | 11 | 0 | 1 | 12 | 7 | 0 |
0 | 2 | 14 | 5 | 10 | 12 | 0 | 0 |
0 | 0 | 6 | 13 | 10 | 0 | 0 | 0 |
1 | 0 | -1 |
1 | 0 | -1 |
1 | 0 | -1 |
-33 | -27 | 14 | 3 | 6 | 27 |
-40 | -10 | -30 | -17 | -11 | 34 |
-35 | 10 | 35 | -26 | -24 | 28 |
-31 | 13 | -30 | -28 | -22 | 29 |
-33 | -6 | 22 | -28 | -4 | 33 |
-31 | -12 | 10 | -6 | 14 | 24 |
4.理解卷积如何提取图像特征。
读取一个图像;
以下矩阵为卷积核进行卷积操作;
显示卷积之后的图像,观察提取到什么特征。
1 | 0 | -1 |
1 | 0 | -1 |
1 | 0 | -1 |
1 | 1 | 1 |
0 | 0 | 0 |
-1 | -1 | -1 |
-1 | -1 | -1 |
-1 | 8 | -1 |
-1 | -1 | -1 |
代码:
from PIL import Image
import matplotlib.pyplot as plt
import numpy as np
from scipy.signal import convolve2d
I = Image.open(r'E:/作业/大三上/人工智能/works/robot/8/test.jfif')
L = I.convert('L')
e = np.array(I) # 原图
eg = np.array(L) # 灰度图
k1 = np.array([[1, 0, -1], [1, 0, -1], [1, 0, -1]]) # 垂直边缘检查
k2 = np.array([[1, 1, 1], [0, 0, 0], [-1, -1, -1]]) # 水平边缘
k3 = np.array([[-1, -1, -1], [-1, 8, -1], [-1, -1, -1]])
p1 = convolve2d(eg, k1, boundary='symm', mode='same')
p2 = convolve2d(eg, k2, boundary='symm', mode='same')
p3 = convolve2d(eg, k3, boundary='symm', mode='same')
I.show()
plt.matshow(p1)
plt.matshow(p2)
plt.matshow(p3)
plt.show()
原图:
![](https://img2020.cnblogs.com/blog/1785418/202006/1785418-20200605215735748-38067382.png)
1 | 0 | -1 |
1 | 0 | -1 |
1 | 0 | -1 |
![](https://img2020.cnblogs.com/blog/1785418/202006/1785418-20200605215848225-1789122512.png)
1 | 1 | 1 |
0 | 0 | 0 |
-1 | -1 | -1 |
![](https://img2020.cnblogs.com/blog/1785418/202006/1785418-20200605215933778-867923402.png)
-1 | -1 | -1 |
-1 | 8 | -1 |
-1 | -1 | -1 |
![](https://img2020.cnblogs.com/blog/1785418/202006/1785418-20200605215944548-2004359216.png)
卷积API
scipy.signal.convolve2d
tf.keras.layers.Conv2D
5. 安装Tensorflow,keras
参考:https://blog.csdn.net/u011119817/article/details/88309256
6. 设计手写数字识别模型结构,注意数据维度的变化。
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten, Conv2D, MaxPool2D
model = tf.keras.Sequential()
model.add(Conv2D(…))
model.add(MaxPool2D(…))
...
#可以上传手动演算的每层数据结构的变化过程。model.summary()
代码:
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D
import os
# 建立模型
model = Sequential()
# 一层卷积
model.add(
Conv2D(
filters=16, # 卷积核种类
kernel_size=(3, 3), # 卷积核大小
padding='same',
input_shape=(28, 28, 1),
activation='relu'))
# 池化层1
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 二层卷积
model.add(
Conv2D(
filters=32,
kernel_size=(3, 3),
padding='same',
activation='relu'))
# 池化层2
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 三层卷积
model.add(
Conv2D(
filters=128,
kernel_size=(3, 3),
padding='same',
activation='relu'))
model.add(Flatten()) # 平坦层
model.add(Dense(128, activation='relu')) # 全连接层
model.add(Dropout(0.5))
model.add(Dense(10, activation='softmax')) # 激活函数
model.summary()
![](https://img2020.cnblogs.com/blog/1785418/202006/1785418-20200606191511054-1828108638.png)
参考:
https://www.jianshu.com/p/afe485aa08ce
https://blog.csdn.net/junjun150013652/article/details/82217571