Added yolov9 compatibility

[Hamzalopode]

Added Yolov9 compatibility:

• Yolov9 model download for E & C
• Yolov9 process ( same as yolov8 )

Other model size and demo predict will be added with time.

[fcakyon]

Amazing contribution @Hamzalopode! 🚀

Please fix the errors raised in tests by:

• Install required development packages:

pip install -e ."[dev]"

• Reformat with black and isort:

python -m scripts.run_code_style format

• Commit and push 💯

[Hamzalopode]

Thanks @fcakyon !

I think everything looks good now. Let me know if there is anything else.

🚀

[fcakyon]

@Hamzalopode for some reason, you updated all the prior notebook files. Can you please revert these changes?

Other than that all seems OK 🚀

[Hamzalopode]

Done.

Sorry I thought that applying the styling format for the notebooks was a good idea.

[GuillaumeBruand]

@Hamzalopode
Can we add one level of abstraction to integrate common features between yolov8 and yolov9 (see e.g. PR #988)
As far as I know there is full compatibility between the two models.

[Hamzalopode]

Fixed merge conflict between the two branch.

[eniktab]

getting the following error when trying to use

TypeError: ERROR ❌️ ../weights/best.pt appears to be an Ultralytics YOLOv5 model originally trained with https://github.com/ultralytics/yolov5.
This model is NOT forwards compatible with YOLOv8 at https://github.com/ultralytics/ultralytics.
Recommend fixes are to train a new model using the latest 'ultralytics' package or to run a command with an official Ultralytics model, i.e. 'yolo predict model=yolov8n.pt'

The training was like

train_dual.py --sync-bn ---img 640 --cfg models/detect/yolov9-c.yaml --weights '' --name yolov9-c --hyp hyp.scratch-high.yaml

[weertman]

Interested in using yolov9 but cannot figure out how to get this to work. I have a custom yolov9-e model and its yaml but I cannot find any documentation on how to use it with SAHI. I've been butting my head against this for a bit and wondering if this has had any progress.

[fcakyon]

@weertman @eniktab @Hamzalopode @GuillaumeBruand

Have you tried this with the latest sahi release? It should be working:

detection_model = AutoDetectionModel.from_pretrained(
    model_type='ultralytics',
    model_path='yolov9e.pt',
    confidence_threshold=0.3,
    device="cpu", # or 'cuda:0'
)

[Hamzalopode]

@weertman @eniktab @Hamzalopode @GuillaumeBruand

Have you tried this with the latest sahi release? It should be working:

detection_model = AutoDetectionModel.from_pretrained(
    model_type='ultralytics',
    model_path='yolov9e.pt',
    confidence_threshold=0.3,
    device="cpu", # or 'cuda:0'
)

This is an old fork (11mo old) when yolov9 just came out and wasn't in sahi at that time, I implemented it for my self-use.

I will take a look again on what I did and see if it's deprecated or already implemented.

[GuillaumeBruand]

@fcakyon @Hamzalopode I am now using sahi with yolov11 and it seems that no additional change is required in the main branch to work properly (just use the yolov8 base class).
To avoid code duplication I recommand refactoring this PR for multiple yolo-vX variations.