基于mmcv的VideoReader获取当前帧的时间戳

首先实例化化一个mmcv的videoreader，传入的参数为视频路径

import mmcv

video_reader = mmcv.VideoReader(video_path)

通过阅读mmcv.VideoReader的源码得知VideoReader的方法其实就是调用了opencv的VideoCapture中的方法。

videocapture简介VideoCapture()中参数是0，表示打开笔记本的内置摄像头，参数是视频文件路径则打开视频，如cap = cv2.VideoCapture("../test.avi")

class VideoReader:

    def __init__(self, filename, cache_capacity=10):
        .....
        self._vcap = cv2.VideoCapture(filename)
        .....

    @property
    def vcap(self):
        """:obj:`cv2.VideoCapture`: The raw VideoCapture object."""
        return self._vcap

mmcv的videoreader中很多方法都是调用这个vcap方法@property的作用，所以我就参考cv2获取时间戳，调用了这个vcap也就是调用cv2的VideoCapture来基于mmcv框架获取时间戳。主要是在使用mmcv框架时比较方便。

video_reader = mmcv.VideoReader(video_path)

for frame in track_iter_progress(video_reader):
    result = inference_detector(model, frame, test_pipeline=test_pipeline)
    # 获取时间戳
    milliseconds = video_reader.vcap.get(cv2.CAP_PROP_POS_MSEC)
    seconds = milliseconds//1000
    milliseconds = milliseconds%1000
    minutes = 0
    hours = 0
    if seconds >= 60:
        minutes = seconds//60
        seconds = seconds % 60

    if minutes >= 60:
        hours = minutes//60
        minutes = minutes % 60

    # print(int(hours), int(minutes), int(seconds), int(milliseconds))
    times = "{}时:{}分:{}秒".format(hours, minutes, seconds)
    #