将UCF101数据集的视频提取为帧（多线程实现）

1.实现代码

最开始的版本是使用opencv实现的，但是在实施的过程中，发现opencv在读取视频帧的时候，有的时候ret, frame = capture.read()中的返回值ret有时会为False，会存为错误的1帧，所以改用ffmpeg实现。
使用多线程的ffmpeg实现代码如下：

# 将UCF101数据集中的视频提取为帧
from __future__ import print_function, division
import os
import sys
import subprocess
from multiprocessing import Pool
from tqdm import tqdm

n_thread = 30

# file_name: v_ApplyEyeMakeup_g01_c01.avi
# class_path: xxx/Videos/ApplyEyeMakeup
# dst_class_path: xxx/jpgs/ApplyEyeMakeup
def vid2jpg(file_name, class_path, dst_dir_path):
    if '.avi' not in file_name:
        return
    name, ext = os.path.splitext(file_name)
    dst_directory_path = os.path.join(dst_dir_path, name)  # dst_directory_path: xxx/jpgs/v_ApplyEyeMakeup_g01_c01

    video_file_path = os.path.join(class_path, file_name)  # video_file_path: xxx/Videos/ApplyEyeMakeup/v_ApplyEyeMakeup_g01_c01.avi
    try:
        if os.path.exists(dst_directory_path):
            if not os.path.exists(os.path.join(dst_directory_path, 'img_00001.jpg')):
                subprocess.call('rm -r \"{}\"'.format(dst_directory_path), shell=True)
                print('remove {}'.format(dst_directory_path))
                os.mkdir(dst_directory_path)
            else:
                print('*** convert has been done: {}'.format(dst_directory_path))
                return
        else:
            os.mkdir(dst_directory_path)
    except:
        print(dst_directory_path)
        return
    cmd = 'ffmpeg -i \"{}\" -threads 1 -vf scale=-1:331 -q:v 0 \"{}/img_%05d.jpg\"'.format(video_file_path, dst_directory_path)
    # print(cmd)
    subprocess.call(cmd, shell=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)


def class_process(dir_path, dst_dir_path, class_name):  # dir_path: xxx/Videos dst_dir_path: xxx/jpgs ApplyEyeMakeup
    print('*' * 20, class_name, '*'*20)
    class_path = os.path.join(dir_path, class_name)  # class_path: xxx/Videos/ApplyEyeMakeup
    if not os.path.isdir(class_path):
        print('*** is not a dir {}'.format(class_path))
        return

    # dst_class_path = os.path.join(dst_dir_path, class_name)  # dst_class_path: xxx/jpgs/ApplyEyeMakeup
    # if not os.path.exists(dst_class_path):
    #     os.mkdir(dst_class_path)

    vid_list = os.listdir(class_path)
    vid_list.sort()
    p = Pool(n_thread)
    from functools import partial
    worker = partial(vid2jpg, class_path=class_path, dst_dir_path=dst_dir_path)
    for _ in tqdm(p.imap_unordered(worker, vid_list), total=len(vid_list)):
        pass
    # p.map(worker, vid_list)
    p.close()
    p.join()

    print('\n')


if __name__ == "__main__":
    dir_path = sys.argv[1]  # /home/xxx/data/projects/datasets/UCF101/Videos
    dst_dir_path = sys.argv[2]  # /home/xxx/data/projects/datasets/UCF101/jpgs

    class_list = os.listdir(dir_path)
    class_list.sort()
    for class_name in class_list:  # class_name: ApplyEyeMakeup
        class_process(dir_path, dst_dir_path, class_name)