Bringing Sports Action to Life in 3D: A New Way to Watch the Game

Sports broadcasts often show athletes moving in a flat, 2D way. This makes it hard to see the true 3D motion of their limbs. To fix this, experts use the lines on sports fields to figure out the camera's position and the scene's geometry. But when cameras zoom in on athletes, those helpful field lines disappear. This makes it tough to get accurate 3D data. A new method called partial sports field registration helps solve this problem. It finds a set of camera positions that work with the scene. By optimizing both the 3D pose of athletes and the camera's position, researchers can now track running movements on a 400m track in 3D. This is a big step forward for sports analysis. To test this method, a large synthetic dataset was created. It includes over 10, 000 images with different viewpoints, running styles, and body types. This dataset shows the limits of current 3D pose estimation methods. The goal is to improve these methods and make them more useful for sports broadcasts. This new approach combines advances in 2D human pose estimation and camera calibration. It opens up new ways to collect large-scale kinematic datasets. These datasets can help improve sports analysis and broadcasting. The synthetic data and code used in this research are available for others to use. This can help more researchers and developers work on improving 3D pose estimation methods.