Arrange Database for ImageGenerator

For train in keras deep learning model we have two options 1) write our generator 2) arrange the images as python expects .

The second options seems to me more interesting

for documentation we can go to https://keras.io/api/preprocessing/image/

Im based on story of https://vijayabhaskar96.medium.com/tutorial-image-classification-with-keras-flow-from-directory-and-generators-95f75ebe5720

so lets write small script to arrange images in this format

Imports

import os
import random
import numpy as np
import shutil

Paths

This is where all classes of images exist , lets say we have 3 classes.

pathClassA = r”pathClassA”

pathClassB = r”pathClassB”

pathClassC = r”pathClassC”

allInputPaths = [pathClassA,pathClassB,pathClassC]
classes = ['pathClassA','pathClassB','pathClassC']

Lets define output folder

outfolder = r"Data"
subFolders = ['Train','Test','Valid']

lets define split of data

splitlen = np.array([0.7 ,0.2 ,0.1])
splitlen[0]  = splitlen[0]
splitlen[1]  = splitlen[0]+splitlen[1]
splitlen[2]  = 1

Main copy loop

mainClass = 0
for name in allInputPaths:
    print(name)
    filelist = os.listdir(name)
    print(len(filelist))
    random.shuffle(filelist)
    n = len(filelist)
    k = 0
    for file_name in filelist:

        full_file_name = os.path.join(name, file_name)

        if (k < n*splitlen[0]):
            dest = os.path.join(os.path.join(outfolder, subFolders[0]), classes[mainClass])
        elif (n*splitlen[0] <= k <= n*splitlen[1]):
            dest = os.path.join(os.path.join(outfolder, subFolders[1]), classes[mainClass])
        else:
            dest = os.path.join(os.path.join(outfolder, subFolders[2]), classes[mainClass])

        if os.path.isfile(full_file_name):
            shutil.copy(full_file_name, dest)
            k = k + 1

    mainClass = mainClass + 1

main folder

Train folder

Thats All folks