split_train_val_test.py

import glob
from os.path import join as opj

root = 'data/MultiLabelAU/normal'
aligned = True

Train_files = sorted(glob.glob(opj(root, '*', 'Training.txt')))
Test_files = sorted(glob.glob(opj(root, '*', 'Test.txt')))

for file_txt in (Train_files):
    print(file_txt.split('/')[-2])
    lines = open(file_txt).readlines()
    train_subj = sorted(list(set([line.split('/')[-3] for line in lines])))
    val_subj = train_subj.pop(int(file_txt.split('/')[-2].split('_')[-1]) + 2)
    print("Train subj: %s" % (str(train_subj)))
    print("Val subj: %s" % (str(val_subj)))

    train_txt = file_txt.replace('Training', 'train')
    val_txt = train_txt.replace('train', 'val')
    test_file = file_txt.replace('Training', 'Test')

    if aligned:
        train_txt = train_txt.replace('normal', 'aligned')
        val_txt = val_txt.replace('normal', 'aligned')

    f_val = open(val_txt, 'w')
    f_train = open(train_txt, 'w')

    for i in range(len(lines)):
        if 'flip' in lines[i]:
            continue
        file_ = lines[i].replace('Codes/ActionUnits/data/Faces',
                                 'datos/Databases')
        if 'BP4D_256' in file_:
            file_ = file_.replace('_256', '')

        file_ = file_.replace('Sequences', 'Faces')
        if aligned:
            file_ = file_.replace('Faces', 'Faces_aligned')

        if val_subj in lines[i]:
            f_val.writelines(file_)
        else:
            f_train.writelines(file_)

    f_train.close()
    f_val.close()

    lines = open(test_file).readlines()
    test_subj = sorted(list(set([line.split('/')[-3] for line in lines])))
    print("Test subj: %s" % (str(test_subj)))
    if aligned:
        test_file = test_file.replace('normal', 'aligned')

    test_file = test_file.replace('Test', 'test')
    f_test = open(test_file, 'w')

    for i in range(len(lines)):
        if 'flip' in lines[i]:
            continue
        file_ = lines[i].replace('Codes/ActionUnits/data/Faces',
                                 'datos/Databases')
        if 'BP4D_256' in file_:
            file_ = file_.replace('_256', '')

        file_ = file_.replace('Sequences', 'Faces')
        if aligned:
            file_ = file_.replace('Faces', 'Faces_aligned')

        f_test.writelines(file_)
    f_test.close()
    print("Len Train: " + str(len(open(file_txt).readlines())))
    print("Len Val: " + str(len(open(val_txt).readlines())))
    print("Len Test: " + str(len(open(test_file).readlines())))
    print("")