The Pose class represents a detected human pose in an image. It encapsulates information about the pose's keypoints, confidence score, bounding box, and unique ID. Additionally, it provides methods for managing pose-related data, such as drawing the pose on an image, updating its ID, and applying smoothing to its keypoints.

Class Definition

class Pose:

The Pose class is designed to store and manage information about a single detected pose.

Class Attributes

These are shared across all instances of the Pose class:

1. num_kpts:

   num_kpts = 18
   

   • The number of keypoints in a pose (e.g., nose, neck, shoulders, etc.).

2. kpt_names:

   kpt_names = ['nose', 'neck', 'r_sho', 'r_elb', 'r_wri', 'l_sho', 'l_elb', 'l_wri',
                'r_hip', 'r_knee', 'r_ank', 'l_hip', 'l_knee', 'l_ank',
                'r_eye', 'l_eye', 'r_ear', 'l_ear']
   

   • Names of the keypoints for easier reference.

3. sigmas and vars:

   sigmas = np.array([...], dtype=np.float32) / 10.0
   vars = (sigmas * 2) ** 2
   

   • sigmas: Standard deviations for each keypoint, used for similarity calculations.
   • vars: Variances derived from sigmas.

4. last_id:

   last_id = -1
   

   • Tracks the last assigned pose ID, ensuring unique IDs for each pose.

5. color:

   color = [0, 224, 255]
   

   • The color used to draw the pose on an image.

Instance Attributes

These are specific to each pose instance:

1. keypoints:
   • A NumPy array of shape (18, 2) representing the coordinates of the pose's keypoints.
   • A value of 1 indicates that a keypoint is not detected.
2. confidence:
   • The confidence score of the detected pose.
3. bbox:
   • The bounding box of the pose, calculated using the get_bbox method.
4. id:
   • A unique ID assigned to the pose.