I'm about to disappear for the holiday period, but a couple of observations:
You can start and stop just one encoder. Check out this example.
When it comes to detection, you should be able to maintain reasonably high framerates so long as the detection isn't running in the camera thread (for example, here - the face boxes are a bit laggy because they don't keep up with the video rate). This kind of approach may not be suitable for all applications, of course.
Having said that, on a Pi 5 both encoder and detection will both be running on the CPU, so unsurprisingly there will be contention there.
You can start and stop just one encoder. Check out this example.
When it comes to detection, you should be able to maintain reasonably high framerates so long as the detection isn't running in the camera thread (for example, here - the face boxes are a bit laggy because they don't keep up with the video rate). This kind of approach may not be suitable for all applications, of course.
Having said that, on a Pi 5 both encoder and detection will both be running on the CPU, so unsurprisingly there will be contention there.
Statistics: Posted by therealdavidp — Wed Dec 18, 2024 6:43 pm
