Pose Estimation Example
July 19, 2026 ยท View on GitHub
Introduction
Pose estimation identifies keypoints such as joints, landmarks, or other distinctive object features.
Usage
To create annotations for a YOLO pose dataset:
- Start by adding the image files.
- Click the
Rectanglebutton on the left toolbar, or pressR, to draw a bounding box and assign itslabelandgroup_id. - Next: Click the
Pointbutton on the left menu bar to draw keypoints on the object. Assign the samegroup_idto the keypoints as the corresponding rectangle to link them together. Keep in mind:- If a keypoint is not visible, you may omit it.
- If a keypoint is obscured, check the
useDifficultfield.
For each object, all associated keypoints and the corresponding rectangle should have the same group_id, which must be unique within the current image.
Tip
- Press
Sto hide selected shapes andWto show hidden shapes. - Use the group ID filter to inspect related shapes. For details, see Shape Display.
- Select multiple shapes and press
Gto automatically assign sequential group IDs; pressUto remove group IDs from all selected shapes.
Export
To export your pose estimation annotations, proceed with the following steps:
- Click on the
Exportbutton located in the menu bar at the top. - Select the
Export YOLO-Pose Annotationsoption. - Upload your custom label file, e.g., pose_classes.yaml, to ensure the correct mapping of object classes.
These steps will facilitate the proper export of your pose estimation annotations for further use or analysis.
Tip
For faster annotation, press Ctrl+Shift+G to enable automatic reuse of the most recent group ID when creating shapes.
To understand the dataset format for YOLO-Pose, refer to the official documentation and consult the sample output file here.