Sending pre-recorded video through Vitamin D

[ComputerUser]

Vitamin D has excellent motion detection capabilities. It would be great if there was a feature to take a pre-existing saved video file and send it through the Vitamin D engine so it can highlight the important parts through its motion detection engine.

Vitamin D uses a lot of CPU power, so in some situations we don't need the motion detection running in real-time 24/7. Therefore, if we had a full recording of video in an area for a year, and we want to review a specific week of video and highlight the important parts, it would be great if we could send that 1 week through the Vitamin D engine. This would prevent having Vitamin D be run in real-time through that entire year of video. For example, if there was a large department store with dozens of video cameras recording everything, it may not be feasible to have Vitamin D process all of that video in real-time.

[doug]

ComputerUser,

We actually experimented with this a bit, but ran into quite a few problems to get it to work well. Specifically:

• The postprocessing was quite slow. Putting a full month of video through the motion detection could take a long time.
• Vitamin D Video is really designed around having date/time/location for a video. That ends up being difficult to get right for imported video.

...that's not to say that the above problems can't be solved, but the user experience was certainly not ideal without a lot of assumptions about how the video was original stored / tagged.

Sorry!

-Doug

[Semper Vaporo]

I have often wondered about doing this, and wondered how to go about it. But your comment about it being slow was a bit of an eye-opener.

If a single processor can handle from 2 to 4 cameras in real-time, then that kind of implies that one processor could handle a data stream at 2 to 4 times that of real-time. Thus a day's worth of video (24 hours) would take from 6 to 12 hours to process! A week's worth would take 42 to 84 hours and a month's worth would take 1 to 2 weeks to process! OUCH!

Multiple processors might reduce that processing time, but processing video is pretty much linear and is difficult to parcel out to multiple processors. So even if doing so cuts the processing time in half, refer to the numbers above and you will see why the "user experience" is not "ideal" (rather boring, actually!)

But even at that, there have been times I would like to feed in a pre-recorded video as though from a camera to have VDV process it, even if done in real-time,

[TennesseeTom]

I would imagine it would be pretty easy to set up VLC to convert any recording into a network stream and then point Vitamin D to it in order to analyze. Granted, this would be "real-time" only but certainly provides a way..