Wrapper around opencv's cv2.VideoCapture to simplify reading video files in python.

In a terminal window run:

conda install pyvideoreader -c ncb

or

pip install pyvideoreader

Open a video file and read frame 100:

from videoreader import VideoReader
vr = VideoReader(video_file_name)
print(vr)  # prints video_file_name, number of frames, frame rate and frame size
frame = vr[100]
vr.close()

Or use a context manger which takes care of opening and closing the video:

with VideoReader(video_file_name) as vr:  # load the video
    frame = vr[100]

Supports slice-syntax: vr[start:end:step]. To iterate over all frames you need to use vr[:]. To read every 100th frame, starting at frame 500 and ending at frame 10000 do this:

for frame in vr[500:10000:100]:
    do_something_with(frame)

Lists, tuples or ranges can also be passed as indices, e.g. vr[(42, 314, 999)].

Note that indexing returns a generator - each frame in the indices is read on demand which saves memory. If you need all frames at once you can convert it to a list list(vr[start:end:frame]).

For compatibility, videoreader can also be used like the underlying cv2.VideoCapture:

ret, frame = vr.read()  # read next frame