I find that the recorded events often start too late and drag on for too long. People are already in view when the recording starts. I would like to either:
- Have it start automatically just before the person comes into view (no motion at start, then the person appears)
- Make it user-configurable (e.g. 10 seconds before detection)
For option 1 above, when the AI discovers a noticeable event, it should back up to the point at which there was no motion at all, and then record until there is no motion again.
Also, the recordings are very heavily “front-weighted” in the sense that almost always, the interesting part of the recording occurs in the first 1/4 to 1/3 of the event with nothing interesting in the more than half of the remaining video. It would be nice to have the AI detect no remaining motion and then just stop the recording.