Allow for parallel processing of chunks

[frueter]

It would be awesome if scene_manager.detect_scenes() could be run with a sub frame range on an existing scene manager instance, so that multiple chunks could be run in parallel while sill referencing the first frame of the clip (e.g. when using ContentDetector). In other words run batches in a way that the resulting stats in the scene maneger instance match the ones you' get if you run the entire clip in one go.
Maybe this is already possible somehow?

I am currently writing a UI for this and would like to speed up the detection as much as possible.

EDIT:
I am realising this is detecting scenes for sub ranges is already possible but the challenge is to make share the scene manager instance across threads. Will keep testing..

[Breakthrough]

Feel free to file a feature request or convert this into an issue, I'm open to supporting this use case. This is a good idea, it's not possible with all detectors, but if we consider doing a two-pass approach for most, then it's fine. It's worth exploring how the API might need to change for this.

[frueter]

Great to hear! I have only played with the ContentDetector so far so can't offer ideas for the other ones, but I was thinking if SceneManager.detect_scenes() could be given a concrete reference frame, then each chunk could explicitly be told to use the first frame from the VideoCapture instance might enable me to use QThreadPool to make this multi threaded quite easily.
If each chunk can be made to output coherent stastics that can be merged in the main thread the I'd be happy.
But I am no preofessional programmer so maybe that is naive?!
Here is ascreen shot of my prototype UI: