How to understand the output structure? could merge so many output layers to one layer like yolov8? about yolov9 HOT 4 OPEN

Yes, it could.
But in actually, I think the output of yolov9 models are aux prediction and main prediction respectively.
You could just choose output of main prediction.

from yolov9.

Yes, it could. But in actually, I think the output of yolov9 models are aux prediction and main prediction respectively. You could just choose output of main prediction.

It means i only param the output0 could get all the objets?

from yolov9.

Yes, it could. But in actually, I think the output of yolov9 models are aux prediction and main prediction respectively. You could just choose output of main prediction.

so what the differents about aux prediction and main prediction, and what aux prediction means?

from yolov9.

I think 2731 would be output of main branch.
Aux branch uses assistant loss to guide training, which is usually used in deep supervision.
In YOLOv9, we applied aux branch in proposed PGI.
Prediction of aux branch often has worse performance than prediction of main branch.

from yolov9.