(python) Convolutional Neural Networks: Step by Step - padding

on Deep Learning

Coursera 참고 예제) Convolutional Neural Network 1주차. Convolutional Neural Networks: Step by Step. CNN구현

Zero padding

image image

import numpy as np
import h5py
import matplotlib.pyplot as plt

%matplotlib inline
plt.rcParams['figure.figsize'] = (5.0, 4.0) # set default size of plots
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'

%load_ext autoreload
%autoreload 2

np.random.seed(1)


# GRADED FUNCTION: zero_pad

def zero_pad(X, pad):  
    X_pad = np.pad(X, ((0,0),(pad, pad), (pad, pad),(0,0)), 'constant', constant_values=0)
                       #      ,(위,아래),(좌,우)  
    return X_pad
    
#임의의 data 만들기 
# batch size = 2
# height = 4
# width = 4
# channel = 3
np.random.seed(1)
x = np.random.randn(2, 4, 4, 3)

# data x에 zero padding 적용
x_pad = zero_pad(x, 2)  

# 시각화
# imshow : array에 색을 채워서 이미지로 보여줌
fig,axarr= plt.subplots(2,4)
axarr[0][0].set_title('x batch1 whole')
axarr[0][0].imshow(x[0,:,:,0]+x[0,:,:,1]+x[0,:,:,2])   
axarr[0][1].set_title('x batch1 channel1')
axarr[0][1].imshow(x[0,:,:,0])  
axarr[0][2].set_title('x batch1 channel1')
axarr[0][2].imshow(x[0,:,:,1]) 
axarr[0][3].set_title('x batch1 channel1')
axarr[0][3].imshow(x[0,:,:,2])

axarr[1][0].set_title('x_pad batch1 whole')
axarr[1][0].imshow(x_pad[0,:,:,0]+x_pad[0,:,:,1]+x_pad[0,:,:,2])  
axarr[1][1].set_title('x_pad')
axarr[1][1].imshow(x_pad[0,:,:,0])
axarr[1][2].set_title('x_pad')
axarr[1][2].imshow(x_pad[0,:,:,1])
axarr[1][3].set_title('x_pad')
axarr[1][3].imshow(x_pad[0,:,:,2])

batch1의 결과
맨 앞이 전체이미지, 뒤에가 채널 각각.
image

batch2의 결과
image

size 비교, batch랑 채널수는 일정하고 크기만 달라짐. image