Development of an autonomous vehicle using deep learning

OData support
Supervisor:
Szemenyei Márton
Department of Control Engineering and Information Technology

In today’s world, computer vision, and deep learning are rapidly developing fields. Tech companies sees the future in these technologies, because they can be used for such problems, that traditional algorithms can not solve. Most notably, they developing in this field for self driving cars, because they see the breakthrough from this technology. In case of a self driving car, natural objects around the vehicle must be properly recognized, because the car could not be able to make good decisions without it. Within this problem, i worked on recognizing road signs. My thesis work was the continuation of developing an autonomous robot, such that it will be able to recognize and localize road signs. For image detection, convolutional neural networks are used, so i solved the task using these. For the localization problem I used the YOLO technology. In my thesis I was able to develop a neural network architecture, which was able to run faster than the small sized TinyYOLO architecture. By developing the project further, it can be achievable to run on embedded systems.

Downloads

Please sign in to download the files of this thesis.