Deployment and Evaluation of Robust Neural Video Compression for Autonomous Driving

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Introduction

Neural video compression is an emerging research field that is quickly gaining traction as
alternative to standardized hand-crafted video codecs. Modern neural codecs are capable of
exceeding performance of existing state-of-the-art standard codecs. Recently real-time
capable neural architectures were introduced, proving the ability to deploy them in real-time
scenarios on modern hardware.

Current neural codecs exhibit vulnerability to transmission errors and lack sophisticated rate
control mechanisms required for practical deployment under variable network conditions.
Packet loss causes catastrophic quality degradation due to temporal prediction
dependencies, while existing rate control approaches fail to maintain consistent quality under
bandwidth fluctuations.

The goal of this work is to extend an existing real-time neural video codec with learned error
correction and rate control mechanisms, implement hierarchical packet

Work Packages:
- Implementation of learned forward error correction using reinforcement learning agents for
adaptive redundancy allocation
- Development of rate control mechanism via deep reinforcement learning to maintain target
bitrate while maximizing perceptual quality
- Design of hierarchical packet coding scheme with progressive quality layers for graceful
degradation
- Integration and deployment on EDGAR vehicle hardware with network simulation for
various channel conditions
- Optimization via TensorRT or comparable acceleration libraries for inference efficiency
- Comparative evaluation against algorithmic rate control methods and standard forward
error correction schemes

1. Nonlinear Transform Coding
2. End-to-End Neural Video Compression: A Review
3. DCVC-RT

4. GRACE

If you are interested or have any questions, please send me an e-mail (niklas.krauss@tum.de) with your CV and a
current transcript of your records, thank you!

 

Voraussetzungen

Requirements:
- Programming experience with Python and well versed with Pytorch, and ideally with a systems programming language (C,C++,Rust, Zig, Odin)
- High personal motivation and independent working style.
- Very good language proficiency in German, English

Verwendete Technologien

Neural Data Compression, Python, Pytorch, Machine Learning, Machine Learning, Super Resolution, Compression, Autonomous Driving, Autonomous Vehicles, Teleoperation, Tensorrt, Forward Error Correction

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