Reorganize image files with imagenet style folder hierarchy

imagenet_poppins.py

import numpy as np
import shutil
from sklearn.model_selection import train_test_split

cats = ['negative', 'positives']

for cat in cats:
    print(cat)
    if not os.path.exists(data_folder + "/train/" + cat):
        os.makedirs(data_folder + "/train/" + cat)
    if not os.path.exists(data_folder + "/test/" + cat):
        os.makedirs(data_folder + "/test/" + cat)
    images = os.listdir(data_folder + "/" + cat)
    # assume that the images have a name like 23431234_Figure2.jpg, do train and test splitting
    train_ids, test_ids = train_test_split(list(set([name.split("_")[0] for name in images])))
    for image in images:
        if image.split("_")[0] in train_ids:
            shutil.copy(data_folder + "/" + cat + "/" + image, data_folder + "/train/" + cat + "/" + image)
        elif image.split("_")[0] in test_ids:
            shutil.copy(data_folder + "/" + cat + "/" + image, data_folder + "/test/" + cat + "/" + image)