@inproceedings{BotschHuberBalasubramanianetal.2023, author = {Botsch, Michael and Huber, Werner and Balasubramanian, Lakshman and Flores Fernandez, Alberto and Geisler, Markus and Gudera, Christian and Morales Gomez, Mauricio Rene and Riegl, Peter and S{\´a}nchez Morales, Eduardo and Chandra Sekaran, Karthikeyan and Weinzierl, Michael}, title = {Data Collection and Safety Use Cases in Smart Infrastructures}, booktitle = {15th International ACM Conference on Automotive User Interfaces: Adjunct Conference Proceedings}, publisher = {ACM}, address = {New York}, isbn = {979-8-4007-0112-2}, doi = {https://doi.org/10.1145/3581961.3609858}, pages = {333 -- 336}, year = {2023}, language = {en} } @thesis{Geisler2022, author = {Geisler, Markus}, title = {Development of robust camera-LiDAR fusion architectures for real-time object detection in 3D}, publisher = {Technische Hochschule Ingolstadt}, address = {Ingolstadt}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:573-35622}, pages = {xii, 129}, school = {Technische Hochschule Ingolstadt}, year = {2022}, abstract = {Camera-LiDAR fusion architectures offer the opportunity to tailor perception systems to the requirements of the respective real-world application. By combining the properties of two modalities, an individual compromise between accuracy, robustness, runtime and system cost can be found. However, this requires a deep understanding of the trade-offs that arise from the chosen architectural design. In this context, this thesis focuses on the development of real-time, and robust camera-LiDAR fusion architectures for 3D object detection. In particular, the concept of generating depth features from both modalities is investigated. The first part of this thesis gives an overview of the state of the art of camera and LiDAR sensor based 3D object detection. A step is made towards a general guideline for the development of camera-LiDAR fusion architectures. The second part focuses on the robustness evaluation. A manual reduction of LiDAR detection performance is proposed to simulate the challenges of poorer point cloud quality in real-world applications of camera-LiDAR fusion architectures. This is used in a within-subject study that analyses the impact of reducing LiDAR detection power and investigates the optimal integration of image data for robustness. The third part of the thesis develops a depth feature based fusion architecture. For this purpose, it uses bird's eye view feature maps generated by both sensors. Subsequently, all proposed methods are evaluated. Finally, the low-level fusion of the camera and LiDAR data is analysed. The quantified performance improvement shows the potential of LiDAR-assisted feature generation in otherwise purely image-based architectural parts.}, language = {en} } @unpublished{ChandraSekaranGeislerRoessleetal.2025, author = {Chandra Sekaran, Karthikeyan and Geisler, Markus and R{\"o}ßle, Dominik and Mohan, Adithya and Cremers, Daniel and Utschick, Wolfgang and Botsch, Michael and Huber, Werner and Sch{\"o}n, Torsten}, title = {UrbanIng-V2X: A Large-Scale Multi-Vehicle, Multi-Infrastructure Dataset Across Multiple Intersections for Cooperative Perception}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2510.23478}, year = {2025}, language = {en} }