Currently, there is considerable demand for large 3D datasets, especially due to the growing interest in autonomous driving and other computer graphics or vision applications. The main goal of the project is to generate a high-quality 3D point cloud, a flexible and scalable geometric representation, from a 2D image. The project is largely divided into two steps: generating point clouds for small shapes based on previous work on PointNet, and using the 3D point cloud for virtual or mixed reality-modeling. With such technology, it is possible to generate massive training data for rare edge cases that can be used for designing perception system of self-driving cars.

I am interested in applying the knowledge from classes, internships, and research experience I acquired during my three years at MIT to real-life applications. I have always been interested in machine learning and computer vision and graphics and want to delve deeper into the field. Through this SuperUROP project, I hope to better understand various techniques of machine learning and how they can be applied to modeling the real world.