AI based Multiview Video Compression for Autonomous Vehicles

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Autonomous vehicles generate huge amounts of sensor data, which is semantically processed by neural networks and classical algorithms to perform the dynamic driving task.

Video data is currently the largest chunk of data transmitted. Recently, neural data compression methods have emerged as a valid alternative to classical methods such as MPEG standards. 

We can go one step further and even leverage information from other cameras at the same time.

Autonomous Vehicles typically have more than one camera with overlapping view points.
Temporally these cameras also capture similar information at different points in time.

 

Goal:

The aim of this study is to improve upon the state of the art neural compression method by leveraging other more than a single camera at the same time to increase overall compression efficiency

Your Role:

• Literature research about current neural data compression methods as well as their evaluation methods
• Literature research about multi-view video compression models .
• Extension of a current implementation of the single view approach to a multi view case
• Training and evaluation of the developed method

This work can also be completed in German! 

Voraussetzungen

- Strong interest & motivation for autonomous driving
- Initiative & independent way of working

- Foundational understanding of statsistics and machine learning
- Programming skills, e.g. Python (Experience with Computer Vision Tasks and Pytorch is a bonus) 

 

If you are interested, please send me a grade sheet with your CV to niklas.krauss@tum.de :)

Verwendete Technologien

Neural Data Compression, Python, Pytorch, Machine Learning, Multiview Machine Learning, Autonomous Vehicles

Tags

FTM Studienarbeit, FTM Krauss, FTM AV, FTM AV Safe Operation, FTM Informatik, FTM Teleoperation

Möglicher Beginn

sofort

Kontakt

Niklas Krauß
Raum: 3507
Tel.: +49172 1736882
niklas.krausstum.de

Ausschreibung