import cv2
cap= cv2.VideoCapture("2653582916.mp4")
while(cap.isOpened()):
    ret, frame = cap.read()
    if ret == False:
        break
print(i)
Output on machine 1

Output on machine 2

System Information

Machine 1

OpenCV = 4.1.1.26

OS = Ubuntu 18.04.4 LTS

16 Core / 32GB RAM

Machine 2

OpenCV = 4.1.1.26

OS = Ubuntu 18.04.3 LTS

8 Core /16GB RAM
frame count will vary depends on machines? please help me out.
          
Please collect more information using videoio debug flags: https://docs.opencv.org/4.1.1/de/db1/group__videoio__registry.html
If you want to report a bug provide complete reproducer (including input data).
Usage questions should go to Users OpenCV Q/A forum: http://answers.opencv.org
          
Video used

http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/BigBuckBunny.mp4
Output on machine 1

14300
Output on machine 2

14308
System Information

Machine 1

OpenCV = 4.1.1.26

OS = Ubuntu 18.04.4 LTS

16 Core / 32GB RAM

[cv2.videoio_registry.getBackendName(b) for b in cv2.videoio_registry.getBackends()]

['FFMPEG', 'GSTREAMER', 'INTEL_MFX', 'V4L2', 'CV_IMAGES', 'CV_MJPEG']

cap.getBackendName()

FFMPEG
Machine 2

OpenCV = 4.1.1.26

OS = Ubuntu 18.04.3 LTS

8 Core /16GB RAM

[cv2.videoio_registry.getBackendName(b) for b in cv2.videoio_registry.getBackends()]

['FFMPEG', 'GSTREAMER', 'INTEL_MFX', 'V4L2', 'CV_IMAGES', 'CV_MJPEG']

cap.getBackendName()

FFMPEG
          
Both numbers seems incorrect - ffprobe returns 14315 in nb_frames:
$ ffprobe -show_streams /home/alalek/Downloads/BigBuckBunny.mp4
BTW, on my machine I see 14308 (i7-6700K - 4 cores/8 threads)
          
@alalek so do you say like the number of frames that we get on different machines are different ?
if so,

why / what is making the particular change in the number of frames produced?
Thanks and Regards,

Do you have different versions of FFmpeg installed on these machines? Looks like duplicate of #15352
Package            Machine 1    Machine 2

ffmpeg-python  0.2.0             0.2.0

opencv-python   4.1.1.26       4.1.1.26
ffmpeg-python 0.2.0 0.2.0
opencv-python package uses builtin FFmpeg. So this extra ffmpeg package is not necessarry (or more worse - they may conflict).
More experiments:
i5-6600 (4 cores / 4 threads) + opencv-python==4.1.1.26 ===> 14312
taskset -c 0 or taskset -c 0,1 doesn't change final result (14312)
          
Hi @alalek,
actual FPS of video which I  used to find frame count, is 24FPS.

But when I read it using VideoCapture,
In Machine 1 :

it is returning FPS as 23.2376 and elapsed time 615.380.
In Machine 2:

FPS 26.157557

elapsed time 546.992977976
can you please clear that depends on hardware even FPS will get a change in VideoCapture?? because of that am I getting the different number of frames in different machines?
Thanks
          
I have the same problem but in different OpenCV versions

In 4.4.0.42  I got more frames than 4.5.4.58

ffmpeg-python==0.2.0
          
on Windows with the latest version of OpenCV i tested used

http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/BigBuckBunny.mp4
import cv2
cap= cv2.VideoCapture("d:/test/BigBuckBunny.mp4")
frame_count = cap.get(cv2.CAP_PROP_FRAME_COUNT)
pos_frames = 14000
cap.set(cv2.CAP_PROP_POS_FRAMES, pos_frames)
while(cap.isOpened()):
    ret, frame = cap.read()
    if ret == False:
        break
print("frame_count:", frame_count)
print("starting frame:", pos_frames)
print("read frame_count:", i)
print("result is:", frame_count - pos_frames - i == 0)
result:
frame_count: 14315.0

starting frame: 14000

read frame_count: 315

result is: True
tested changing pos_frames and always result is: True
          
Same here. I did not see any problem until I upgrade opencv to version 4.5.5.64.

I built it for Windows and for Linux and the problem appears only on Windows. I have a video containing 609 frames and cap.get(cv2.CAP_PROP_FRAME_COUNT) returns 611 on Windows, whereas it correctly returns 609 frames on Linux. Before upgrading both returned 609 frames.
Is there any progress on this issue?
          
I was experiencing incorrect frame counts with MP4 videos specifically. People having having trouble with MP4 videos in specific seems like a common trend in this thread. For what it's worth, here are some outputs counting the frames of the same video only difference being the video format (one more frame for mov is odd):