Directed Research Presentation
3D scene analysis framework for Self Driving Cars
The project focuses on developing a 3D scene analysis framework for the development of self driving cars. The frame work involves 3D Mapping and Localization (SLAM), 2D Semantic Segmentation, 3D semantic Mapping, 2D/3D detection and tracking of the traffic participants in a traffic scenario. The application for this framework is to serve as a platform for motion planning, motion prediction algorithms and traffic behavior understanding. Current state of the art algorithms are used for each module. In the presentation I would like to highlight the challenges in some of these modules. The future work will be to develop a probabilistic framework for joint 2D-3D detection and tracking. The research was conducted at Honda Research Institute, Mountain View, USA.
Advisor: Professor Loris Fichera