Publications

@inproceedings{11097781,

title = {BEV-LLM: Leveraging Multimodal BEV Maps for Scene Captioning in Autonomous Driving},

author = {Felix Brandstätter and Erik Schütz and Katharina Winter and Fabian B. Flohr },

url = {https://ieeexplore.ieee.org/abstract/document/11097781},

doi = {10.1109/IV64158.2025.11097781},

isbn = {979-8-3315-3803-3},

year  = {2025},

date = {2025-08-06},

booktitle = {2025 IEEE Intelligent Vehicles Symposium (IV)},

pages = {345-350},

publisher = {IEEE},

abstract = {Autonomous driving technology has the potential to transform transportation, but its wide adoption depends on the development of interpretable and transparent decision-making systems. Scene captioning, which generates natural language descriptions of the driving environment, plays a crucial role in enhancing transparency, safety, and human-AI interaction. We introduce BEV-LLM, a lightweight model for 3D captioning of autonomous driving scenes. BEV-LLM leverages BEVFusion to combine 3D LiDAR point clouds and multi-view images, incorporating a novel absolute positional encoding for view-specific scene descriptions. Despite using a small 1B parameter base model, BEV-LLM achieves competitive performance on the nuCaption dataset, surpassing state-of-the-art by up to 5% in BLEU scores. Additionally, we release two new datasets — nu-View (focused on environmental conditions and viewpoints) and GroundView (focused on object grounding) — to better assess scene captioning across diverse driving scenarios and address gaps in current benchmarks, along with initial benchmarking results demonstrating their effectiveness.},

howpublished = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{10.1145/3581961.3609844,

title = {Using Node-RED as a Low-Code Approach to Model Interaction Logic of Machine-Learning-Supported EHMIs for the Virtual Driving Simulator Carla},

author = {Sven Winkelmann and Max Büttner and Dharani Deivasihamani and Alexander Hoffmann and Fabian Flohr},

url = {https://doi.org/10.1145/3581961.3609844},

doi = {10.1145/3581961.3609844},

isbn = {9798400701122},

year  = {2023},

date = {2023-01-01},

booktitle = {Adjunct Proceedings of the 15th International Conference on Automotive User Interfaces and Interactive Vehicular Applications},

pages = {323–326},

publisher = {Association for Computing Machinery},

address = {Ingolstadt, Germany},

series = {AutomotiveUI '23 Adjunct},

abstract = {External Human-Machine Interfaces (eHMI) enable interaction between vehicles and Vulnerable Road Users (VRU), for example, to warn VRUs of the car’s presence. Warning systems should warn of the situation’s urgency, which can be achieved using Machine Learning (ML)-based VRU detection models. ML models and eHMI interaction concepts are usually developed by different teams and tested separately, often resulting in integration problems. This work contributes to a low-code approach to model interaction concepts involving ML models to enable end-to-end prototypes for early integration and User eXperience (UX) testing. We use flow-based modeling with Node-RED, the virtual driving simulator CARLA and YOLOv5 as state-of-the-art deep learning techniques for VRU detection. We show two scenarios (cornering lights and context-aware VRU warning) in an interactive demonstrator, meaning a manual live control of pedestrian and car. We consider our approach to model and evaluate interaction concepts without writing code feasible for non-computer scientists.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{pmlr-v151-triess22a,

title = {Point Cloud Generation with Continuous Conditioning},

author = {Larissa T Triess and Andre Bühler and David Peter and Fabian B Flohr and Marius Zöllner},

editor = {Gustau Camps-Valls and Francisco J R Ruiz and Isabel Valera},

url = {https://proceedings.mlr.press/v151/triess22a.html},

year  = {2022},

date = {2022-03-01},

booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics},

volume = {151},

pages = {4462--4481},

publisher = {PMLR},

series = {Proceedings of Machine Learning Research},

abstract = {Generative models can be used to synthesize 3D objects of high quality and diversity. However, there is typically no control over the properties of the generated object.This paper proposes a novel generative adversarial network (GAN) setup that generates 3D point cloud shapes conditioned on a continuous parameter. In an exemplary application, we use this to guide the generative process to create a 3D object with a custom-fit shape. We formulate this generation process in a multi-task setting by using the concept of auxiliary classifier GANs. Further, we propose to sample the generator label input for training from a kernel density estimation (KDE) of the dataset. Our ablations show that this leads to significant performance increase in regions with few samples. Extensive quantitative and qualitative experiments show that we gain explicit control over the object dimensions while maintaining good generation quality and diversity.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{braun2021simple,

title = {Simple Pair Pose-Pairwise Human Pose Estimation in Dense Urban Traffic Scenes},

author = {Markus Braun and Fabian Berthold Flohr and Sebastian Krebs and Ulrich Kressel and Dariu M Gavrila},

year  = {2021},

date = {2021-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

pages = {1545--1552},

organization = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{zhi2021pose,

title = {Pose-Guided Person Image Synthesis for Data Augmentation in Pedestrian Detection},

author = {Rong Zhi and Zijie Guo and Wuqiang Zhang and Baofeng Wang and Vitali Kaiser and Julian Wiederer and Fabian B Flohr},

year  = {2021},

date = {2021-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

pages = {1493--1500},

organization = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang2021urbanpose,

title = {UrbanPose: A New Benchmark for VRU Pose Estimation in Urban Traffic Scenes},

author = {Sijia Wang and Diange Yang and Baofeng Wang and Zijie Guo and Rishabh Kumar Verma and Jayanth Ramesh and Christoph Weinrich and Ulrich Kressel and Fabian Berthold Flohr},

year  = {2021},

date = {2021-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium Proceedings},

pages = {1537--1544},

organization = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{chandan2021posessd,

title = {VRU Pose-SSD: Multiperson Pose Estimation For Automated Driving},

author = {Chandan Kumar and Jayanth Ramesh and Bodhisattwa Chakraborty and Renjith Raman and Christoph Weinrich and Anurag Mundhada and Arjun Jain and Fabian B Flohr},

year  = {2021},

date = {2021-01-01},

booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},

volume = {35},

number = {17},

pages = {15331--15338},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{guo2020generativeb,

title = {Generative Model Based Data Augmentation for Special Person Classification},

author = {Zijie Guo and Rong Zhi and Wuqiang Zhang and Baofeng Wang and Zhijie Fang and Vitali Kaiser and Julian Wiederer and Fabian B Flohr},

year  = {2020},

date = {2020-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{guo2020generative,

title = {Generative Model based Data Augmentation for Special Person Classification},

author = {Rong Zhi and Zijie Guo and Wuqiang Zhang and Baofeng Wang and Vitali Kaiser and Julian Wiederer and Fabian Berthold Flohr},

year  = {2020},

date = {2020-01-01},

booktitle = {2020 IEEE Intelligent Vehicles Symposium (IV)},

pages = {1675--1681},

organization = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{zhijie2020traffic,

title = {Traffic Police Gesture Recognition by Pose Graph Convolutional Networks},

author = {Zhijie Fang and Wuqiang Zhang and Zijie Guo and Rong Zhi and Baofeng Wang and Fabian B Flohr},

year  = {2020},

date = {2020-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{wang2019leverage,

title = {Leverage of Limb Detection in Pose Estimation for Vulnerable Road Users},

author = {Sijia Wang and Fabian B Flohr and Hui Xiong and Tuopu Wen and Baofeng Wang and Mengmeng Yang and Diange Yang},

year  = {2019},

date = {2019-01-01},

booktitle = {IEEE International Conference on Intelligent Transportation Systems (ITSC)},

pages = {528--534},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{xiong2019recurrent,

title = {Recurrent Neural Network Architectures for Vulnerable Road User Trajectory Prediction},

author = {Hui Xiong and Fabian B Flohr and Sijia Wang and Baofeng Wang and Jianqiang Wang and Keqiang Li},

year  = {2019},

date = {2019-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

pages = {171--178},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{krebs2017survey,

title = {A survey on leveraging deep neural networks for object tracking},

author = {Sebastian Krebs and Bharanidhar Duraisamy and Fabian B Flohr},

year  = {2017},

date = {2017-01-01},

booktitle = {IEEE International Conference on Intelligent Transportation Systems (ITSC)},

pages = {411--418},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{aparicio2017advancing,

title = {Advancing active safety towards the protection of vulnerable road users: the PROSPECT project},

author = {Andrés Aparicio and Laura Sanz and Gary Burnett and Hans Stoll and Maxim Arbitmann and Martin Kunert and Fabian B Flohr and Patrick Seiniger and Dariu M Gavrila},

year  = {2017},

date = {2017-01-01},

booktitle = {International Technical Conference on the Enhanced Safety of Vehicles (ESV)},

publisher = {National Highway Traffic Safety Administration},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{li2016new,

title = {A new benchmark for vision-based cyclist detection},

author = {Xiaofei Li and Fabian B Flohr and Yue Yang and Hui Xiong and Markus Braun and Shuyue Pan and Keqiang Li and Dariu M Gavrila},

year  = {2016},

date = {2016-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

pages = {1028--1033},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{roth2016driver,

title = {Driver and pedestrian awareness-based collision risk analysis},

author = {Markus Roth and Fabian B Flohr and Dariu M Gavrila},

year  = {2016},

date = {2016-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium (IV)},

pages = {454--459},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{braun2016pose,

title = {Pose-rcnn: Joint object detection and pose estimation using 3d object proposals},

author = {Markus Braun and Qing Rao and Yikang Wang and Fabian B Flohr},

year  = {2016},

date = {2016-01-01},

booktitle = {IEEE International Conference on Intelligent Transportation Systems (ITSC)},

pages = {1546--1551},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{kooij2014context,

title = {Context-based pedestrian path prediction},

author = {Julian F P Kooij and Nicolas Schneider and Fabian B Flohr and Dariu M Gavrila},

year  = {2014},

date = {2014-01-01},

booktitle = {European Conference on Computer Vision (ECCV)},

pages = {618--633},

publisher = {Springer},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{flohr2014joint,

title = {Joint probabilistic pedestrian head and body orientation estimation},

author = {Fabian B Flohr and Madalin Dumitru-Guzu and Julian F P Kooij and Dariu M Gavrila},

year  = {2014},

date = {2014-01-01},

booktitle = {IEEE Intelligent Vehicles Symposium Proceedings (IV)},

pages = {617--622},

publisher = {IEEE},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{flohr2013pedcut,

title = {PedCut: an iterative framework for pedestrian segmentation combining shape models and multiple data cues},

author = {Fabian B Flohr and Dariu M Gavrila},

year  = {2013},

date = {2013-01-01},

booktitle = {British Machine Vision Conference (BMVC)},

publisher = {BMVA},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{fischer2012evaluation,

title = {Evaluation of tracking methods for maritime surveillance},

author = {Yvonne Fischer and Marcus Baum and Fabian B Flohr and Uwe D Hanebeck and Jürgen Beyerer},

year  = {2012},

date = {2012-01-01},

booktitle = {SPIE Defense, Security, and Sensing},

pages = {839208--839208},

publisher = {International Society for Optics and Photonics},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}