-
Notifications
You must be signed in to change notification settings - Fork 42
Description
We have been operating the camera at 15.0 Hz on the highest power mode the Orin can run at (0). We would like to operate the jetson at the next highest power mode (1), but tests shows that this may not be possible due to either how we've implemented the wrapper around the driver or due to actual camera/VimbaCPP api limitations.
I was wondering if this is a bug or expected behavior on jetson platforms? In continuous mode, could this be narrowed down to the FrameObserver callback? Some analysis i've run suggests this could be an issue with this capture thread in continuous mode. What i've observed is that logs immediately at the beginning of the frame callback, we are receiving images at exactly 15Hz, but we basically copy the frame off the capture thread towards the end of the callback (note: there is some processing in between. But profiling shows latency that shouldn't cause issues at this rate), and what ends up on our side is not the same that we see w/i the frame callback. We're getting images at ~13.4Hz on our side (what is copied off the thread and published), but ==15Hz w/i the frame callback. Wondering if any of this sounds familiar to those in the know.
Let me know if any additional information may be helpful. Thanks!