Nerd For Tech
Published in

Nerd For Tech

Case Study of Keypoint Annotation

ByteBridge Key Points Annotation

Keypoint Annotation

Keypoint labeling, including those of face, human body, and objects of specific types, is often used to train face recognition models and statistical models.

The face annotation with keypoints is to accurately locate the facial features by measuring the distance between the eyes, the shape of the chin, the distance between the nose and the mouth, and generate the so-called “fingerprint” of the face, which is suitable for the uniqueness of a specific individual code.

Keypoint annotation of the human body is to track and monitor the human body (hands, arms, head, and legs) for online sports teaching, such as yoga. In the driving process, the driver can be detected when they are on a phone call or smoking. In the security field, it can be used to monitor workers’ abnormal behaviors that are not consistent with the safety rules.

Labellers need to have a three-dimensional sense in keypoint annotation so that the labelled image can be stereoscopic. When there are obscured points or invisible points, the labeler should have the ability of spatial imagination in the process.

Keypoint Labelling Requirements

Here are the labeling names and the order. And it is essential to proceed according to the ID order in the above figure.

Nose, left eye, right eye, left ear, right ear, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left hip, right hip, left knee, right knee, left ankle, and right ankle.

Labelling Instructions

Keypoints need to be labeled in the center of the joint.

Predict and label the image even if you cannot see specific key points such as eyes and ears.

With certain joints unseen, label them if you can predict their position.

Note: Only if you cannot see and predict them, there is no need for labeling.

Challenges in Keypoint Annotation

First of all, there are usually a lot of key points. And in the labeling process, one needs to figure out their meaning, which can be several or even hundreds in quantity. Therefore, the more the key points are, the bigger the chance to make mistakes.

Secondly, the criteria over whether the labeling is good or not is vague, resulting in less chance for labelers to review mistakes. Given so many labeled points, a point deviated from the location is not easy to find in time. After all, people are not machines and for the naked eye to find an equal diversion point between two points is challenging.

You Configure and We Annotate

Outsource your data labeling tasks to ByteBridge, you can get high-quality ML training datasets cheaper and faster!

  • Free Trial Without Credit Card: you can get your sample result in a fast turnaround, check the output, and give feedback directly to our project manager.
  • 100% Human Validated
  • Transparent & Standard Pricing: clear pricing is available(labor cost included)

Why not have a try?



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store



Data labeling outsourced service: get your ML training datasets cheaper and faster!—