Feature Request: Detect/Drop static scenes with no change in video contet

[nutsinu]

I do not believe this is currently supported with PySceneDetect. It would be useful to implement a feature to detect scenes without significant change (specified by a threshold) in video content. This way we can drop non useful scenes (e.g. static background or image)

[wjs018]

This would be possible to do with certain detectors using the stats and some kind of filter. However, another project by @Breakthrough might be better suited for this purpose. Check out DVR-Scan! Sounds like it might be more what you are looking for.

In general, something replicating this kind of behavior could be useful for certain applications. I will add it to my backlogged list of things to explore.

[Breakthrough]

This is definitely on the radar for v1.0. Currently in the v1.0 branch, each detector now returns an event type (cut, in, or out), and the plan is to allow use of both detector types at the same time. Thus it should be possible to combine a threshold/motion detector (which will output in/out events) so that the output only contains the "zones" those detectors find (i.e. the output scenes will only be those between IN and OUT events).

That can additionally be combined with the content/adaptive detectors which output cut events, so that those areas between adjacent in/out events can be further split with CUT events. As @wjs018 pointed out, DVR-Scan might be better suited to your purposes for now, although eventually the plan is to have better integration between those two tools (and have more overlap) once the v1.0 refactor is completed.

[zaterio]

#318