create dataset

create_dataset.py

import os
import cv2
from shutil import copyfile
from random import randint,shuffle
import string


def create_dir(path):
    if not os.path.exists(path):
        os.makedirs(path)

def create_dirs(path):
    vowels = "अ आ इ ई उ ऊ ए ऐ ओ औ अं अ:".split()
    consonants = "क ख ग घ ङ च छ ज झ ञ ट ठ ड ढ ण त थ द ध न प फ ब भ म य र ल व श ष स ह क्ष त्र ज्ञ".split()
    nums = [ str(i) for i in range(0, 10) ]
    dirs = vowels + consonants + nums
    for d in dirs:
        create_dir(path + '/' + d)

def create_pathmap():
    root = 'ncr-original/'
    dst = 'data-keras/'
    create_dirs(dst + 'train')
    create_dirs(dst + 'val')

    classes = os.listdir(root)
    pathmap = {}

    # dtype => vowels, consonants, nums
    for dtype in classes:
        # labels => क, ख, etc...
        labels = os.listdir(root + dtype)
        for label in labels:
            img_path = root + dtype + '/' + label + '/'
            for image in os.listdir(img_path):
                load_path = img_path + image
                try:
                    l = pathmap[label]
                except KeyError:
                    pathmap[label] = []
                pathmap[label].append(load_path)
    return pathmap

def create_dataset2(pathmap, train=0.8, val=0.2):
    root = 'ncr-original/'
    dst = 'data-keras/'
    dst_train = dst + 'train'
    dst_val = dst + 'val'

    create_dirs(dst_train)
    create_dirs(dst_val)

    for label in pathmap:
        load_paths_all = pathmap[label]
        images_num = len(load_paths_all)
        train_idx = int(images_num * train)
        val_idx = int(images_num * val)
        load_paths_train = load_paths_all[:train_idx]
        load_paths_val = load_paths_all[train_idx : train_idx + val_idx]

        print(label, images_num, len(load_paths_train), len(load_paths_val))

        # train set
        for img_path in load_paths_train:
            fname = str(randint(0,999999999999))
            img = cv2.imread(img_path)
            write_path = dst_train + '/' + label + '/' + fname + '.jpg'
            print(write_path)
            cv2.imwrite(write_path, img)

        # val set
        for img_path in load_paths_val:
            fname = str(randint(0,999999999999))
            img = cv2.imread(img_path)
            write_path = dst_val + '/' + label + '/' + fname + '.jpg'
            print(write_path)
            cv2.imwrite(write_path, img)

def main():
    pathmap = create_pathmap()
    create_dataset2(pathmap, train=0.8, val=0.2)


if __name__=='__main__':
    main()