Detection bounding box data

[IanOMHiggins]

Expose Detection mode bounding boxes' bounds and labels in plain text rather than only RTSP overlay.

[IanOMHiggins]

(VIO payload)

Is there a reason why this data is not exposed already? I just want the bounds and labels, which already appear clearly in the overlay [see attached image, the orange box around person with "ID:1 person 92.7%" in red]. All I want is the string "ID:1 person 92.7%" as well as the bounds of the orange box without needing to extract it from the image directly.

I made a ticket with Gremsy support regarding this. They said it was "impossible" because of "technical limitations and the potential impact on real-time operation", but it's just a few bytes of already-computed data. This seems to imply it was a design decision rather than an oversight or a work-in-progress.

I understand that there's a time-sync aspect to this, and there are questions about how much of the feature, label, and confidence data to include, but there's several ways to overcome that. It couldn't be any worse than the current condition, where the entire object detection framework is essentially useless despite clearly functioning properly.

Talking with Gremsy support was quite frustrating.

Below you can see that the RTSP stream shows the bounding boxes and labels (when debug overlay and detection mode are enabled).

[imgodofmoney]

Hi bro,
Please send an email to me at duong.nguyen@gremsy.com
I want to know more detail about your request to see how can we help you in this case.