Integració d'informació geogràfica al motor Unity3D

Abstract

Nowadays deep learning is been used more and more for many autonomous driving applications, however for effectively using deep learning we need a lot of special annotations called ground truth. Annotate this data manually is very hard and maybe impossible. That's why the easiest way to get this huge quantity of data is to use graphical engines like Synthia does (using Unity3D). However, to generate a synthetic city like this you need also a lot of manual work, modeling, texturing and architecture of the city and the terrain and that requires a lot of labor. For this work we propose to do a system that is included inside the Synthia project that from different kinds of geographic information that is obtained from Internet, we are capable of generating in real time a 3D approximation of the world and from here we can extract data from the same quality as those that are currently generated.

Date of Award	28 Jun 2017
Original language	Catalan
Supervisor	Antonio Manuel Lopez Peña (Director)