Samasource-1: How to create an annotation system for 3D point cloud data using an immersive experience(vr)

LIDAR sensors paired to Machine learning models is a standard for self driving cars, it is now also becoming mainstream with mobile devices, customer appliances and gaming devices, often to enhance camera data with 3D information about its environment. To build a Machine learning model from LIDAR data, training datasets have to be created. This is currently done via manual annotation of point cloud data by Annotators in a 2D desktop application. This project is aiming at solving the complexity of 3D annotation by using a VR based application where an agent is immersed into the 3D world captured by the LIDAR sensor. The operator is able to navigate in this world as if they were part of the scene, through the VR application and draw bounding boxes around the various objects identified in the scene.