Why model inference result shape is 1 84 8400? about ultralytics HOT 4 CLOSED

@dongqianzhuan where are you seeing this shape? Can you please provide a minimal reproducible code? Thanks!

from ultralytics.

@dongqianzhuan where are you seeing this shape? Can you please provide a minimal reproducible code? Thanks!
i run this code

`from ultralytics import YOLO

model = YOLO('yolov8s.pt')

model.export(format="onnx")`

i export onnx ,and watch it in netron,it`s dim different with yolov5
if i want run it with my yolov5_onnx_inference project,i need transpose yolov8_res(0,2,1)
i want to know why use this new shape ,it is more efficient?

from ultralytics.

@dongqianzhuan if I remember this correctly, we removed some unnecessary torch.permute operation. Then it's (1, 84, 8400).

from ultralytics.

OK i understand,thank u

from ultralytics.