The yolov5n segmentation example for Hailo -8L with RPi 5 works perfectly fine. It properly detects and segments humans, cars, buses, bikes and everything.
But if I want to only segment the human, can someone please tell me how do I do it?
The yolov5n segmentation example for Hailo -8L with RPi 5 works perfectly fine. It properly detects and segments humans, cars, buses, bikes and everything.
But if I want to only segment the human, can someone please tell me how do I do it?
Hi @shubham,
There are two options here:
Regards,
Hi @shubham,
To be more specific - the change in the code should be applied here:
Where you iterate over the detection only if the label is the relevant one (I think the label here would be “person”).
Regards,
Hello @Omer
Thank you for the prompt reply. The application seems to be filtering “person” providing the relevant information. This is useful! Thanks.
About the first approach that you mentioned, retraining the model, do we have complete documentation for it?
Hi @shubham,
Basically, this is not a Hailo tool as Hailo provides the framework for inference and not training. This is something you can probably do using Ultralytics.
However, for certain models Hailo does provide a retraining docker. You can look at the available models here:
Specifically in your case, a retraining docker for yolov5_seg doesn’t exist in our Model Zoo, but you can use yolov8_seg.
Regards,
That’s even better! I was anyways looking for yolov8_seg solutions.
Hello @Omer,
I was able to get the hef for my retrained model. But I am stuck at this step to get the custom post process library object. I am not able to find how to generate this .so.
Can you please help me with this?
Hi @shubham,
The example you are trying to run is specific for yolov5_seg. You can use the example for yolov8_seg here:
Regards,
Hi @Omer
I tried to use the example from Hailo-Application-Code repo with Hailo-8L. I got the attached error when trying to run instance segmentation with yolov8seg
Hi @shubham,
The error you are getting is exactly what’s described here:
Did you tun the the steps in the above like when installing Hailo on your rpi?
Regards,