“…When it comes to commercialization, we always try to reduce the cost and improve the time efficiency on the condition that the good performance is still maintained. Thanks to the recent advances on deep learning based algorithms regarding computer vision, cameras have shown their capability in the object detection [7,8], tracking [9], and localization [3,10,1]. In autonomous driving, it is a critical task to exploit the potential of off-the-shelf monocular cameras for reliable object 3D localization in the real-world road scenes, which can be a powerful addition of outside driving environment awareness, with or without the availability of Radar or LiDAR.…”