OpenCV loading video in the memory

Question

when using cv::VideoCapture to read video from the disk, what does exactly happens?

1. Reading the whole file in the memory then decoding it frame by frame?
2. Fetch frame by frame directly for the disk?

If the second option is the answer, is it possible to do it in the first way?

Answer 1

To accomplish this in Python on Linux, I simply made a copy of the input file to /dev/shm, which is a RAM-backed filesystem.

tempdir = tempfile.TemporaryDirectory(dir='/dev/shm')
copypath = os.path.join(tempdir.name, os.path.basename(args.video))
shutil.copy(args.video, copypath)
args.video = copypath

Of course there is the cost of the copy operation as well but it shouldn't be much more than simply reading the entire file (since there is no slow disk that's being written to on the other side of the copy).