We are using the term pose detection here, but this is also known as pose estimation. If you have not come across pose estimation, this simply refers to a computer vision operation where human figures are detected, either from an image or from a video. Once the figure(s) have been detected, the model is able to determine roughly where the key joints and body segments (such as the left ear) are.

The growth of pose detection has been rapid, and it has some obvious usages. For instance, we could use pose detection to perform motion capture for animation; studios are increasingly turning to motion capture to capture live-action performances and convert them into 3D images. Another usage lives in the field of sports; in fact, sports have many potential usages of motion capture. Suppose you are a pitcher in a major league baseball team. Pose detection could be used to determine whether your stance was correct at the point of releasing the ball; perhaps you were leaning over too far, or your elbow positioning was incorrect. With pose detection, it becomes easier for coaches to work with players to correct potential problems.

At this point, it's worth noting that pose detection is not the same as person recognition. I know it seems obvious, but there are people who have been confused by this technology into thinking that this somehow identified who a person is. That's a completely different form of machine learning.