make_data_list.py

"""Script for creating Google Speech Commands reduced label dataset"""
import json
from argparse import ArgumentParser
from utils.dataset import get_train_val_test_split
import os


def main(args):
    pretrain_list, train_list, val_list, test_list, label_map = get_train_val_test_split(args.data_root,
                                                                                         args.val_list_file,
                                                                                         args.test_list_file,
                                                                                         args.pretrain)
    if not os.path.isdir(args.out_dir):
        os.makedirs(args.out_dir)

    if len(pretrain_list) > 0:
        with open(os.path.join(args.out_dir, "pretraining_list.txt"), "w+") as f:
            f.write("\n".join(pretrain_list))

    with open(os.path.join(args.out_dir, "training_list.txt"), "w+") as f:
        f.write("\n".join(train_list))

    with open(os.path.join(args.out_dir, "validation_list.txt"), "w+") as f:
        f.write("\n".join(val_list))

    with open(os.path.join(args.out_dir, "testing_list.txt"), "w+") as f:
        f.write("\n".join(test_list))

    with open(os.path.join(args.out_dir, "label_map.json"), "w+") as f:
        json.dump(label_map, f)

    print("Saved data lists and label map.")


if __name__ == "__main__":
    parser = ArgumentParser()
    parser.add_argument("-v", "--val_list_file", type=str, required=True, help="Path to validation_list.txt.")
    parser.add_argument("-t", "--test_list_file", type=str, required=True, help="Path to test_list.txt.")
    parser.add_argument("-d", "--data_root", type=str, required=True,
                        help="Root directory of speech commands v2 dataset.")
    parser.add_argument("-o", "--out_dir", type=str, required=True,
                        help="Output directory for data lists and label map.")
    parser.add_argument("--pretrain", type=float, required=False, default=None,
                        help="Amount of train set reserved for pretraining, e.g., 0.8 for 80%.")
    args = parser.parse_args()

    main(args)