why cant i change the bounding box on the primary window [hailo detection app window]

General
1

Hi.
I am trying to switch from Nvidia to Hailo and having few confusion
I just want to detect one object [example only car ]
when I am doing it, the program open a new window [User Frame]- which I think is unnecessary and also resource consuming and this new window is not projecting in full screen and few buttons are always on top.
I am unable to find any documentation
any way to run ultracytics over hailo ?
any help well be appreciated
thanks

2

Hey @shahnawaz_hamid,

Regarding the user frame - this is intended for writing on top of frames in Python and cannot be executed directly. Please do not use the --user-frame flag.

We have compiled most YOLO models in our model zoo with straightforward compilation procedures. You can access them at: GitHub - hailo-ai/hailo_model_zoo: The Hailo Model Zoo includes pre-trained models and a full building and evaluation environment

For single object detection, you can modify the labels JSON file accordingly.

3

Dear Omria

Greetings

the issue is not of detection but of the primary window.
i can not change or add the logic in the primary window and if i have to
add my logic in the window then a new window appear
that to never in the centre and with lots of button on the top.
cant we just add our own bounding box logic on the main window ?

thanks

4

Hi.

great, i will not use --user-frame flag.
the issue is - how to make basic changes
like instead of square box i want circle
or i want filled box or any other info on top of box
where is the sample code for these kind of things.
i havent found any code to write any thing

thanks

5

Hey @shahnawaz_hamid,

I see what you’re trying to do - you want to add custom drawing on top of the video frames, right?

Unfortunately, this isn’t as straightforward as it should be with our current setup. The main display window you see when running HailoRT CLI is controlled by the hailooverlay GStreamer element. There’s no Python hook into that window, so you can’t directly add your own drawings like circles or custom labels without modifying the plugin source code.

Good news is we’re working on making this easier in future releases.

For now, you’ve got a few workarounds:

Option 1: Separate window approach
Use a user-frame callback to grab the frame data, do your OpenCV drawing, then display it in your own window alongside the main one.

Option 2: Custom pipeline
Write your own Python/GStreamer script that skips the hailooverlay entirely. Use the same pipeline elements but pull frames back to Python for custom drawing, then show with cv2.imshow.

Option 3: Overlay on overlay
Create your own overlay with the Python drawing you want and pipe it after the hailooverlay element.

The user-frame callback (Option 1) is probably your quickest path if you don’t mind having two windows open.

If you want to dig deeper into the source, check out hailo-tappas-core where these plugins live.

6

Thanks Omria

for your response
i am using user-frame but there is a fairly good amount of delay -
[lagging]
of around 2 -3 seconds
for other two options can you plesase send some link

Thanks

7

Hey @shahnawaz_hamid,

For option 2, we don’t actually have an example like that available. But basically, you’d be handling the drawing of the boxes and any other graphics you want through your own overlay implementation.

As for option 3, I’ve personally tried using cairooverlay right after hailooverlay and it worked pretty well for me - I was able to write text and draw on the frames without any issues.

Just a heads up though - we’re currently working on developing a better way to handle this, so keep an eye out for updates!

Hope that helps clarify things!