The conf parameter setting in the pose task. about ultralytics HOT 1 OPEN

Hello! In YOLOv8 pose estimation tasks, individual keypoint confidence thresholds cannot be set directly via the model's configuration or command-line interface. Instead, the model uses a single confidence threshold for all keypoints.

To set a global confidence threshold for all keypoints, you can modify the conf parameter when calling the predict method, similar to other detection tasks. Here's an example using the Python interface:

from ultralytics import YOLO

# Load your model
model = YOLO('yolov8n-pose.pt')

# Run predictions with a global confidence threshold
results = model.predict('path/to/image.jpg', conf=0.25)

This sets a confidence threshold of 0.25 for all keypoints detected in the image. Adjust the conf value as needed to meet your specific requirements. 🌟

from ultralytics.