pip install supervision roboflow
version=rf.workspace("roboflow-jvuqo").project("football-field-detection-f07vi").version(8)
version.download("yolov8")
kp_model=version.modelassertkp_modelisnotNonefirst_image_path=os.listdir(f"{dataset.location}/train/images")[0]
first_image_path=f"{dataset.location}/train/images/{first_image_path}"result=kp_model.predict(first_image_path).json()
# Failskeypoints=sv.KeyPoints.from_inference(result)
Tested on several projects: