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 Rectangle button on the left toolbar, or press R, to draw a bounding box and assign its label and group_id.
  • Next: Click the Point button on the left menu bar to draw keypoints on the object. Assign the same group_id to 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 useDifficult field.

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 S to hide selected shapes and W to show hidden shapes.
  • Use the group ID filter to inspect related shapes. For details, see Shape Display.
  • Select multiple shapes and press G to automatically assign sequential group IDs; press U to remove group IDs from all selected shapes.

Export

To export your pose estimation annotations, proceed with the following steps:

  1. Click on the Export button located in the menu bar at the top.
  2. Select the Export YOLO-Pose Annotations option.
  3. 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.