predict.py file

I’m trying to train a YOLOv8 model on my custom dataset. Each time I export the model, the output shape is (1, 15, 8400). Since I have 11 classes, the expected output shape should be (x, y, w, h, objectness score, 11 labels). With the same dataset, when I train a YOLOv5 model, it correctly outputs (1, 25200, 16). Why is YOLOv8 giving an unexpected output shape?

Here is my code in python :
`
from ultralytics import YOLO
model = YOLO(‘best.pt’)
results = model.predict(
source = ‘/Users/sereentaleb/Desktop/download.jpg’,
save=True,
conf=0.25
)

Im using YOLO (v8.2.25) and when I run the code below it won’t work:

Searching in GitHub and in the code I see contradictory statements. Does the normalization is just dividing the image intensity values (in range [0, 255]) by 255 (so it’s in float range [0,1]) or it does normalization using z-score with some values (e.g. ImageNet mean and std as is suggested here)?

From image bellow (taken from here), we can see that bbox output in yolov8 has size 4 x reg_max. In yolov8 implementation, the reg_max is set to 16 (16 predicted bboxes) so the output has size 64.

i want to use default YOLOv8 model (yolov8m.pt) for object detection. I know that default YOLO models uses COCO dataset and can detect 80+ objects. I just want to detect like 5 of them, how can i achieve this?