New AI Model Boxer Converts 2D Object Detection to 3D Bounding Boxes

Researchers introduced Boxer, a transformer-based algorithm that lifts 2D object detections into 3D bounding boxes using depth data and posed images, requiring less costly 3D training data. The approach significantly outperforms existing methods on benchmark tests, advancing open-world 3D object localization for robotics and autonomous systems.