Larissa Triess

PhD Student · Deep Generative Models · Domain Adaptation · Autonomous Driving · Stuttgart, Germany

Hi there! I'm Larissa, a PhD student working on autonomous driving at Mercedes-Benz AG. I'm also a part of the reseach group of Prof. Marius Zöllner at Karlsruhe Institute of Technology (KIT). My research is in the field of machine learning and 3D vision with focus on synthetsizing realistic LiDAR point clouds for domain adaptation applications.


Publications

Papers

A Survey on Deep Domain Adaptation for LiDAR Perception
IEEE Intelligent Vehicles Symposium (IV) 2021, Workshop Autonomy@Scale
Larissa T. Triess, Mariella Dreissig, Christoph B. Rist, J. Marius Zöllner
Scalable systems for automated driving have toreliably cope with an open-world setting. This means, theperception systems are exposed to drastic domain shifts, likechanges in weather conditions, time-dependent aspects, or geographic regions. Covering all domains with annotated data is impossible because of the endless variations of domains and the time-consuming and expensive annotation process. Furthermore, fast development cycles of the system additionally introduce hardware changes, such as sensor types and vehicle setups, and the required knowledge transfer from simulation. To enable scalable automated driving, it is therefore crucial to address these domain shifts in a robust and efficient manner. Over the last years, a vast amount of different domain adaptation techniques evolved. There already exists a number of survey papers for domain adaptation on camera images, however, a survey for LiDAR perception is absent. Nevertheless, LiDAR is a vital sensor for automated driving that provides detailed 3D scans of the vehicle’s surroundings. To stimulate future research, this paper presents a comprehensive review of recent progress in domain adaptation methods and formulates interesting research questions specifically targeted towards LiDAR perception.
 pdf  project
Scan-based Semantic Segmentation of LiDAR Point Clouds: An Experimental Study
IEEE Intelligent Vehicles Symposium (IV) 2020
Larissa T. Triess, David Peter, Christoph B. Rist, J. Marius Zöllner
Autonomous vehicles need to have a semantic understanding of the three-dimensional world around them in order to reason about their environment. State of the art methods use deep neural networks to predict semantic classes for each point in a LiDAR scan. A powerful and efficient way to process LiDAR measurements is to use two-dimensional, image- like projections. In this work, we perform a comprehensive experimental study of image-based semantic segmentation architectures for LiDAR point clouds. We demonstrate various techniques to boost the performance and to improve runtime as well as memory constraints.
 pdf  project  code
CNN-based synthesis of realistic high-resolution LiDAR data
IEEE Intelligent Vehicles Symposium (IV) 2019
Larissa T. Triess, David Peter, Christoph B. Rist, Markus Enzweiler, J. Marius Zöllner
This paper presents a novel CNN-based approach for synthesizing high-resolution LiDAR point cloud data. Our approach generates semantically and perceptually realistic results with guidance from specialized loss-functions. First, we utilize a modified per-point loss that addresses missing LiDAR point measurements. Second, we align the quality of our generated output with real-world sensor data by applying a perceptual loss.
In large-scale experiments on real-world datasets, we evaluate both the geometric accuracy and semantic segmentation performance using our generated data vs. ground truth. In a mean opinion score testing we further assess the perceptual quality of our generated point clouds. Our results demonstrate a significant quantitative and qualitative improvement in both geometry and semantics over traditional non CNN-based up-sampling methods
 pdf  poster  project

Patents

DE102021002684
L.T. Triess, C.B. Rist. 2021. Method for semantic segmentation of first sensor data of a first sensor type.
German Patent DE102021002684. Filed 21.05.2021. Patent pending.
DE102021002689
L.T. Triess, D. Peter. 2021. Method for transforming sensor data.
German Patent DE102021002689. Filed 21.05.2021. Patent pending.
DE102021002559
L.T. Triess, D. Peter. 2021. Method for generating realistic maps of beam outages in simulated LiDAR data.
German Patent DE102021002559. Filed 17.05.2021. Patent pending.
L.T. Triess, D. Peter. 2021. Method for automatic detection and localization of anomalies in data acquired by means of a LiDAR sensor.
German Patent DE102021001043A1. Filed 26.02.2021. Issued 15.04.2021.
L.T. Triess, D. Peter. 2021. Method for training a neural network of an electronic computing device of a motor vehicle.
German Patent DE102021000803A1. Filed 16.02.2021. Issued 15.04.2021.
L.T. Triess. 2020. Method for transforming acquired sensor data from a first data domain into a second data domain.
German Patent DE102020001541A1. Filed 09.03.2020. Issued 01.10.2020.
L.T. Triess, D. Peter. 2020. Method for processing LiDAR sensor data.
German Patent DE102019003621A1. Filed 23.05.2019. Issued 02.01.2020.

Experience

PhD Student

Mercedes-Benz AG · Stuttgart, Germany

Cutting edge research for environment perception in intelligent vehicles. Domain adaptation with deep generative models for 3D LiDAR data. Publication of papers at top research conferences. Commitment in publicly subsidized research projects.

December 2018 - Present

Student Employee · Deep Learning

Daimler AG · Stuttgart, Germany

Master thesis project on synthesis of realistic high-resolution 3D LiDAR point clouds. Accompanied by University of Stuttgart as part of the Master's study program.

October 2018 - April 2018

Research Fellow · Computer Vision

Tokyo Institute of Technology · Tokyo, Japan

Research project about prostate cancer detection in MR-images with deep learning. Exchange student during the Master's study program.

October 2017 - February 2018

Student Employee · Environment Perception

Daimler AG · Stuttgart, Germany

Part-time employment at the Environment Perception unit in the R&D Department. Lane detection on reflectivity provided by 3D LiDAR sensors with traditional and deep learning based methods. Many diverse tasks with usage of Linux, Git, Python, C++, unit tests, and more within a Scrum software development team.

October 2016 - October 2017

Research Fellow · Computer Vision

Voxar Labs · Recife, Brazil

Visiting research student at Voxar Labs, Universidade Federal de Pernambuco. Object detection and tracking in camera images in real-world traffic scenes. Trying not to get robbed with hardly any knowledge of Portuguese.

April 2016 - September 2016

Student Employee · Automotive Networks

Robert Bosch GmbH · Stuttgart, Germany

Development of an media independent interface for PHY configuration implemented in C and Python on a Raspberry Pi. Demonstration of AVB functions and time synchronization in an Ethernet based switched network.

October 2015 - February 2016

Teacher Assistant

University of Stuttgart · Stuttgart, Germany

Employment at several institutes at the University during my Bachelor studies. Responsibilities usually involved preparing and conducting tutorials and lab courses, as well as correction and grading of assignments.

October 2015 - February 2016

Education

Karlsruhe Institute of Technology

Doctor of Philosophy (PhD)

Computer Science - Machine Learning

Fellow in the Leadership Talent Academy, a career building program funded by KIT and Karl Schlecht Foundation (KSG)
December 2018 - Present

University of Stuttgart

Master of Science

Electrical Engineering and Information Technology - Information and Communication Technology

Volunteering as a mentor in the Cross-Cultural Mentoring Program for visiting international students.
Exchange Semesters:
  • Tokyo Institute of Technology, Tokyo, Japan - October 2017 - February 2018
  • Universidade Federal de Pernambuco, Recife, Brazil - April 2016 - September 2016
April 2016 - October 2018

University of Stuttgart

Bachelor of Science
Electrical Engineering and Information Technology - Communication Systems and Signal Processing
Mentee in the Junior Mentoring Program for Bachelor students in technology.
Volunteering as a mentor in the Cross-Cultural Mentoring Program for visiting international students.
October 2012 - March 2016

Activities

Competitions

In my spare time I participate in some coding competitions, like the Google Competitions Code Jam, Kick Start, and Hash Code. I also like to solve the IEEE SB Passau Advent Calendar each year in December.

Google Hash Code 
2020: with a new team strategy we reached place 824 out of 10,724 - find our code on GitHub
2019: we failed hard the first time as a team and made place 5,075 out of 6,640
ZEISS and KIT Robotics Innovation Competition (2018) 
A 24-hour hackathon in teams of four where we had to solve the cup game with a camera and a robot arm. We implemented a detection and tracking algorithm for our three cups and the ball. After shuffling the cups, we control the robot arm via ROS to lift the cup with the ball underneath.

Awards & Certifications

  • Coursera: Graph Search, Shortest Paths, and Data Structures [certificate]
  • Coursera: Divide and Conquer, Sorting and Searching, and Randomized Algorithms [certificate]
  • International Computer Vision Summer School (ICVSS) 2019
  • University of Stuttgart: Intercultural Competence and Internationality [certificate]
  • Baden-Württemberg Scholarship 2017
  • 1st Place in the category "Automotive Technologies" - Women-STEM-Award 2017
  • CNPq Research Scholarship 2016
  • students@bosch - alumni network for top 5% students that worked at Robert Bosch GmbH