图像分类数据集按照比例划分成训练集和验证集

标注文件为txt格式，文件中单行为一条标注记录，将其按照比例划分成为训练集和验证集

训练指令：python divide.py --path [your file path] --train_ratio 0.8

程序运行完以后，会在当前文件所在统计目录下生成trian.txt、和val.txt两个文件

import argparse
import numpy as np


def divide(lines, train_ratio):
    fp_val = open('./val.txt', 'w+', encoding='utf-8')
    fp_train = open('./train.txt', 'w+', encoding='utf-8')
    length = len(lines)

    shuffled_indices = np.random.permutation(length)
    train_size = int(length * train_ratio)
    train_list = shuffled_indices[:train_size]
    val_list = shuffled_indices[train_size:]

    for i in val_list:
        fp_val.write(lines[i])

    for i in train_list:
        fp_train.write(lines[i])
    print('train images ', train_size)
    print('val images ', length - train_size)
    fp_val.close()
    fp_train.close()


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--path', type=str, default='./file.txt')
    parser.add_argument('--train_ratio', type=float, default=0.8)
    args = parser.parse_args()
    fp = open(args.path, 'r', encoding='utf-8')
    lines = fp.readlines()
    divide(lines, args.train_ratio)