Skip to content
BENIBIX Learn

Pose Estimation

Also known as Human Pose Estimation, Keypoint Detection, Skeleton Tracking

Finding a body's joints in a picture. Join the dots for a stick figure.

Think of it like

It is like a stick man drawn over a photo. Dots for joints, lines for arms and legs.

Example

Emma's fitness app watches her squat through the phone camera. It tracks her hips and knees and tells her when her knees cave in.

Why it matters

If software knows how a body is placed, it can check form or read a wave. Robots also use it to move around people.

Definition

Pose estimation finds the key points of a body in an image or video. For a person, these are spots like the nose, shoulders, elbows, hips and knees. A common set uses 17 points. Joining the dots gives a stick figure that shows how the body is placed. It works on video too, so it can follow a dance or a workout. Some models find points on hands or animals instead.
In the middle, dots mark the joints. Lines join them into a stick figure.

Common mistake

Thinking it knows who the person is. It finds joints, not faces, and two people can give the same stick figure.

Key takeaways

  1. 1
    Key points. Nose, shoulders, elbows, knees and more.
  2. 2
    Stick figure. Joined dots show the pose.
  3. 3
    Not who. It finds a pose, not a name.