Alright, it took an embarrassing amount of time to finally get it but apparently all I had to do was add ‘–use-frame’ (e.g. python basic_pipelines/detection.py --use-frame) to add a playback display with the frames that are modified by cv2.
to prevent having two displays I changed this in hailo_apps_infra/gstreamer_app.py:
self.video_sink = “fakesink” # changed from ‘autovideosink’
So now when I run the script I have one display with cv2 drawing/overlaying things into the playback video (along with the bounding boxes), which is exactly what I was looking for to continue developing!
These posts were key for me to figure this out:
Imho this should be the default configuration, running rpi5 examples with cv2 calls doing nothing on the playback is confusing when starting off, even though the solution is fully integrated. Maybe I missed some documentation… Anyway, happy to have found the solution and hopefully this helps others!