Web Application: 第４回画像データをnpyデータに変換 - 流行の科学技術を勝手に追いかけるブログ人工知能仮想現実ブロックチェーン

こんにちはKeita_Nakamori(´・ω・`)です。

前回flickrから取得した画像データをTensor Flowが読めるように数値データに変換していきます。

Contents

必要なモジュールをインストール

pip install Pillow :
pip install scikit-learn :

Pillow-6.1.0とscikit-learn-0.21.3が入りました

スクリプト：generate_inputdata.py

from PIL import Image # image operation
import os             # get file list
import glob           # image treatment
import numpy as np    # culculation
from sklearn.model_selection import train_test_split # split data

# initialize parameters
classes = ['car', 'motorbike'] # define 2 classses 
num_classes = len(classes)     # the number of classes
image_size = 150               # pixels of width or height

# read images and cnvert numpy array
X = [] # ready to make list
Y = [] # ready to make list

# numbering and 
for index, class_label in enumerate(classes):
    
    # create class_label directories (car or motorbike)
    photos_dir ='./' + class_label

    # search jpeg fileb and create files object
    files = glob.glob(photos_dir + '/*.jpg')

    for i, file in enumerate(files):

        # open image files as instance
        image = Image.open(file)

        # convert image into RGB value data
        image = image.convert('RGB')

        # align to the same size (just in case)
        image = image.resize((image_size, image_size))

        # convert RGB value into numpy array 
        data = np.asarray(image)

        # normalize
        data = data / 255.0

        # append value to list
        X.append(data)
        Y.append(index)

# convert list into numpy array
X = np.array(X)
Y = np.array(Y)

# split data (training and test)
X_train, X_test, y_train, y_test = train_test_split(X, Y)

# replace 1 variable
xy =(X_train, X_test, y_train, y_test)

# save values as npy file 
np.save('./image_files.npy', xy)

from PIL import Image # image operation

import os # get file list

import glob # image treatment

import numpy as np # culculation

from sklearn.model_selection import train_test_split # split data

# initialize parameters

classes = ['car', 'motorbike'] # define 2 classses

num_classes = len(classes) # the number of classes

image_size = 150 # pixels of width or height

# read images and cnvert numpy array

X = [] # ready to make list

Y = [] # ready to make list

# numbering and

for index, class_label in enumerate(classes):

# create class_label directories (car or motorbike)

photos_dir ='./' + class_label

# search jpeg fileb and create files object

files = glob.glob(photos_dir + '/*.jpg')

for i, file in enumerate(files):

# open image files as instance

image = Image.open(file)

# convert image into RGB value data

image = image.convert('RGB')

# align to the same size (just in case)

image = image.resize((image_size, image_size))

# convert RGB value into numpy array

data = np.asarray(image)

# normalize

data = data / 255.0

# append value to list

X.append(data)

Y.append(index)

# convert list into numpy array

X = np.array(X)

Y = np.array(Y)

# split data (training and test)

X_train, X_test, y_train, y_test = train_test_split(X, Y)

# replace 1 variable

xy =(X_train, X_test, y_train, y_test)

# save values as npy file

np.save('./image_files.npy', xy)

実行しましょう

(djangoai) C:\Users\keita\anaconda_projects\djangoai>python generate_inputdata.py

フォルダを見ると新しく「image_files.npy」　　609,516,317 （６０９MB？）のデータが生成されました。

次回

Web Application: 第５回　はじめてのwebアプリ

この記事を書いた人
最新の記事

月	火	水	木	金	土	日
				1	2	3
4	5	6	7	8	9	10
11	12	13	14	15	16	17
18	19	20	21	22	23	24
25	26	27	28	29	30

Web Application: 第４回　画像データをnpyデータに変換

必要なモジュールをインストール

スクリプト：generate_inputdata.py

実行しましょう

次回

Keita N

最新記事 by Keita N (全て見る)