Simon Speth

Simon Speth, M.Sc.

Technische Universität München

Informatik 4 - Lehrstuhl für Software und Systems Engineering (Prof. Pretschner komm.)

Postadresse

Tel.: +49 (89) 289-17836
Raum: 5611.01.059
E-Mail: simon.speth@tum.de

About Me

I am a Ph.D. student at the Chair of Software and Systems Engineering headed by Professor Pretschner. My main research focus is on testing machine learning and deep learning models. Specifically, I am interested in a technique called metamorphic testing.

Thesis Topics

If you are interested in writing a thesis related to my research interests, please contact me by email.

Open Topics:

Title	Type
Generation and Analysis of Metamorphic Test Cases for Neural Networks	Any
Scalable Clustering of Traffic Scenarios Without a Mental Model (with F. Huber)	Master

Assigned Topics:

Title	Type
Scalable Clustering of Traffic Scenarios Without a Mental Model	Master

Finished Topics:

Title	Type
Implementing Metamorphic Relations for Testing Deep Learning Systems with the GeMTest Framework	Master
Measuring Bias in Machine Learning Models with Defect-based Metamorphic Testing	Master
Effectiveness of Search-Based Testing on a Deep Reinforcement-Learned Swarm Controller	Bachelor
Improving Code, Documentation, and Test Quality of the Metamorphic Testing Framework	Bachelor
Applying Mutation Testing Tools for a Quality Evaluation of Metamorphic Relations	Bachelor
Evaluating an Auto Encoder-Based Test Case Generation Algorithm for Deep Learning Models	Master
Enhancing Performance of Radar-based Indoor Presence Detection by Targeted Fault-Injected Data Augmentation	Master
Evaluating the Quality of Metamorphic Test Suites for Deep Learning Models with Mutation Testing	Master
Model Repair and Training methods in Deep Learning Models	Guided Research
Implementation of Metamorphic Relations for Testing Open Source Large Language Models	Master
Fixing Faults in Neural Networks by Retraining with Metamorphic Relations	Master
Improvements of a Python-Based Metamorphic Testing Framework	Bachelor
Definition of Abstraction Levels for Metamorphic Testing	Master
Defect Hypothesis Based Metamorphic Testing of 3D Object Detection Systems	Master
Data Augmentation in the Latent Space for Boosting Performance on Radar-Based Presence Sensing Applications	Master
Improving Robustness of Semantic Segmentation for Autonomous Driving	Master
Latent Space-Based Test Case Generation of Naturally Occurring Environmental Conditions for Traffic Sign Classifiers	Master
Inverse Transparency for Cloud Architectures	Bachelor
Metamorphic Testing of LiDAR/RADAR Obstacle Detection Systems	Master
Search-Based Robustness Testing for Deep Learning Computer Vision Systems	Master

Teaching

Semester	Course
Summer 2025	Advanced Topics of Software Testing
Winter 2024/2025	Advanced Testing of Deep Learning Models: Towards Robust AI
Summer 2024	Advanced Topics of Software Testing Advanced Testing of Deep Learning Models: Towards Robust AI
Winter 2023/2024	Advanced Testing of Deep Learning Models: Towards Robust AI
Summer 2023	Advanced Topics of Software Testing
Winter 2022/2023	Advanced Testing of Deep Learning Models: Towards Robust AI
Summer 2022	Advanced Topics of Software Testing Advanced Python Programming
Winter 2021/2022	Introduction to Informatics I From Sub-Systems to Systems of Systems – Developing Autonomous Driving Functions (in cooperation with fortiss) Seminar: Software Quality
Summer 2021	Requirements Engineering Requirements Engineering (Elite program) Seminar: Software Quality

Publications

Unable to fetch resource from https://mediatum.ub.tum.de/services/export/node/1096728/allchildren?q=speth with exception: cURL error 7: Failed to connect to mediatum.ub.tum.de port 443 after 1 ms: Couldn't connect to server (see https://curl.haxx.se/libcurl/c/libcurl-errors.html) for https://mediatum.ub.tum.de/services/export/node/1096728/allchildren?q=speth

Speth, Simon; Trien, Maximilian; Kufer, Dominik; Pretschner, Alexander: Safety-Critical Oracles for Metamorphic Testing of Deep Learning LiDAR Point Cloud Object Detectors. IEEE Open Journal of Intelligent Transportation Systems 6, 2025, 95-108 mehr…
Speth, Simon; Pretschner, Alexander: GeMTest: A General Metamorphic Testing Framework. 2025 IEEE/ACM 47th International Conference on Software Engineering: Companion Proceedings (ICSE-Companion), 2025, 41-44 mehr…
Speth, Simon; Jasper, Christoph; Jordan, Claudius; Pretschner, Alexander: Generating Latent Space-Aware Test Cases for Neural Networks using Gradient-Based Search. Proceedings of the 18th IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW), 2025, 1-8 mehr…
Speth, Simon; Gonçalves, Artur; Rigault, Bastien; Suzuki, Satoshi; Bouazizi, Mondher; Matsuo, Yutaka; Prendinger, Helmut: Deep learning with RGB and thermal images onboard a drone for monitoring operations. Journal of Field Robotics 39 (6), 2022 mehr…