MOG2 vs CNT

[ceewanna]

After testing both methods, it seems CNT always produces at least 1 video event for every input file. This happens regardless what threshold or kernel is set. My test set both CNT and MOG2 gets triggered at the same value of threshold.

I left the CNT on one night (which usually should produces almost no motion video) but there was every motion video for every input file. Then I tested this specifically to a clean input video and found it produces redundant output as outlined above.

I have no idea whether this is the nature of CNT or not.

[Breakthrough]

Hi, do these always occur at the same point in time? E.g. is it always at the beginning of the input file? If you could provide some output logs or samples, that would be very helpful to finding the issue.

Thanks!

[Breakthrough]

I will look into this some more, thank you. It would also help to see the output of -mo mask.avi for the first few seconds of videos where this happens - would you be willing to share that / see if anything interesting happens there? Thanks!

[Breakthrough]

I think I see the issue, indeed this is a bug. Thanks for letting me know, will turn this into a bug report and have it fixed in the next release.

[ceewanna]

I just tested it with -mo on one of my files. I saw one square white mask once at the beginning and the the rest masks seem to follow the path of the moving object in my video.

[ceewanna]

On another file, where there is only 1 event (9.2 sec) detected, I saw a long period of black screen but no one white square happened. The long period of black screen at the beginning seems to be similar to the one I mentioned above. So I guess the one brink at the beginning may be related to the real event in my video. Not so sure if all these help you identify the problem. Let me know if you need anything more.

[Breakthrough]

I filed #119 to follow up on this in the future. I see this occasionally too, so may have to start ignoring the first frame output of CNT to compensate. Thanks for the report!