尝试从 opencv 中的 .mov 视频读取帧时“找不到起始编号（以文件名）”

Question

When running the command cv.VideoCapture(path) (with cv being an opencv-python import)

I receive the following error:

OpenCV(4.5.1) C:\Users\appveyor\AppData\Local\Temp\1\pip-req-build-kh7iq4w7\opencv\modules\videoio\src\cap_images.cpp:253: 
error: (-5:Bad argument) CAP_IMAGES: can't find starting number (in the name of file):
res\Confetti_Loop.mov in function 'cv::icvExtractPattern'

I have tested it on 3 different computers (all Windows 10) (with same imports) and only one ran the following code with no errors.

"""Separate thread file to deal with videos."""

from queue import Empty

import cv2 as cv
import pygame

from Constants import LOCAL_IO


def get_video(path, size, queue, max_queue, aspect=0):
    """Get video and add to queue once properly formatted."""

    video = cv.VideoCapture(path)
    if aspect == 0:
        size = keep_aspect_ratio(video, *size)
    elif aspect == 1:
        size = cut(video, *size)
    else:
        size = (int(size[0]), int(size[1]))
    while True:
        try:
            get = queue.get(False)
        except Empty:
            pass
        else:
            if get[0] == LOCAL_IO["Stop"]:
                break

        ret, frame = video.read()
        if not ret:
            video.set(2, 0.0)
            ret, frame = video.read()

        frame = cv.cvtColor(frame, cv.COLOR_BGR2RGB)
        frame = cv.resize(frame, size)
        frame = frame.swapaxes(0, 1)
        surface = pygame.surfarray.make_surface(frame)
        surface.set_colorkey((0, 0, 0))
        max_queue.put((LOCAL_IO["Video"], surface))


def keep_aspect_ratio(vid, bx, by):
    """Scales image to fit into bx/by, this method will retain the original image's aspect ratio."""

    ix = vid.get(cv.CAP_PROP_FRAME_WIDTH)
    iy = vid.get(cv.CAP_PROP_FRAME_HEIGHT)

    if ix > iy:  # fit to width

        scale_factor = bx / float(ix)
        sy = scale_factor * iy
        if sy > by:
            scale_factor = by / float(iy)
            sx = scale_factor * ix
            sy = by
        else:
            sx = bx
    else:  # fit to height

        scale_factor = by / float(iy)
        sx = scale_factor * ix
        if sx > bx:
            scale_factor = bx / float(ix)
            sx = bx
            sy = scale_factor * iy
        else:
            sy = by

    return int(sx), int(sy)


def cut(vid, bx, by):
    """Scales image without changing aspect ratio but instead growing to at least the size of box bx/by."""

    ix = vid.get(cv.CAP_PROP_FRAME_WIDTH)
    iy = vid.get(cv.CAP_PROP_FRAME_HEIGHT)

    if bx / float(ix) > by / float(iy):  # fit to width

        scale_factor = bx / float(ix)
        sy = scale_factor * iy
        sx = bx

    else:  # fit to height

        scale_factor = by / float(iy)
        sx = scale_factor * ix
        sy = by

    return int(sx), int(sy)

Why is this happening?

Edit:

Seems to be a problem with the file I am trying to read as I re-downloaded the video from my github repo and it worked. The video file was uploaded using git lfs would this be a reason why it got corrupted during the first download of the whole repo?

Answer 1

According to Olli-Pekka Heinis on https://github.com/opencv/opencv-python/issues/284 . He says:

Only reason I can think of is that PyInstaller does not copy the FFmpeg dll during the packaging process.

If you have a look at eg ...Python37\Lib\site-packages\cv2 folder, there is something like opencv_videoio_ffmpeg412_64.dll next to cv2.cp37-win_amd64.pyd.

Does the opencv_videoio_ffmpeg412_64.dll file exist in the package which was created in PyInstaller? If not, add it to the spec file (binaries): https://pyinstaller.readthedocs.io/en/stable/spec-files.html#spec-file-operation