2D to 3D annotation is a method of generating 3D annotations (e.g., 3D bounding boxes) based on 2D image annotations (e.g., 2D bounding boxes, key points) combined with spatial rules, reducing pure 3D annotation costs, used for 3D object detection model training.
