119. 3d bounding box parameterization using vision based cues for autonomous driving
Name: Ishan Gupta
Grad Year: 2018
The current advances in the field of deep learning and object detection architectures has helped us in learning the 2D semantics of the objects in our environment. In case of autonomous driving, we gather information from a lot of sensors like camera, lidar, radar etc. to gather information about the 2D and 3D semantics of our environment. In our work, we are trying to focus on learning complete computer vision based schemes to remove the dependence on complete lidar based architectures. We are learning 3D semantics of the nearby vehicles like the center/depth of the vehicle in the world frame and the real dimensions from a monocular image in an end to end fashion.