Training an Open-Vocabulary Monocular 3D Object Detection Model without 3D Data Yan Wang