Team
CAOR
Biography
Sascha Hornauer is a researcher whose work lies at the intersection of artificial intelligence, multimodal perception, and autonomous systems. His research notably explores the integration of acoustic and visual signals to enhance understanding and interaction with the environment, as evidenced by his contributions to hybrid methods such as NeRAF, which combines radiance and acoustic fields for 3D scene synthesis. His work also addresses practical applications in robotics and autonomous driving, where he develops innovative approaches for motion planning, novelty detection in sensory data, and the use of bio-inspired echolocation for navigation in degraded conditions. Part of his research focuses on optimizing deep learning models, particularly for complex tasks like semantic segmentation, depth prediction, or sound event classification, often leveraging strategies such as knowledge transfer or self-supervised learning. His contributions also include methodological advances for processing real and simulated data, with an emphasis on the efficiency and robustness of autonomous systems.
Publication(s)
-
2025
NERAF: 3D SCENE INFUSED NEURAL RADIANCE AND ACOUSTIC FIELDS
-
2024
PANO-ECHO: PANOramic depth prediction enhancement with ECHO features DOI : 10.1109/CAI59869.2024.00193
-
2024
MBAPPE: MCTS-Built-Around Prediction for Planning Explicitly DOI : 10.1109/IV55156.2024.10588457
-
2023
Instance segmentation of soft-story buildings from street-view images with semiautomatic annotation DOI : 10.1002/eqe.3805
-
2023
GRI: General Reinforced Imitation and Its Application to Vision-Based Autonomous Driving DOI : 10.3390/robotics12050127
-
2023
The Audio-Visual BatVision Dataset for Research on Sight and Sound DOI : 10.1109/IROS55552.2023.10341715
-
2022
Safety Monitoring of Neural Networks Using Unsupervised Feature Learning and Novelty Estimation DOI : 10.1109/TIV.2022.3152084
-
2022
Self-Supervised Feature Learning and Phenotyping for Assessing Age-Related Macular Degeneration Using Retinal Fundus Images DOI : 10.1016/j.oret.2021.06.010
-
2021
Unsupervised discriminative learning of sounds for audio event classification DOI : 10.1109/ICASSP39728.2021.9413482
-
2020
BatVision: Learning to See 3D Spatial Layout with Two Ears DOI : 10.1109/ICRA40945.2020.9196934
-
2020
Scene Novelty Prediction from Unsupervised Discriminative Feature Learning DOI : 10.1109/ITSC45102.2020.9294451
-
2019
Driving scene-retrieval by example from large-scale data
-
2019
Imitation learning of path-planned driving using disparity-depth images DOI : 10.1007/978-3-030-11021-5_33
-
2018
Synchronizing Multiple Data Streams by Time for Vehicle Control DOI : 10.1109/IVS.2018.8500470
-
2013
Towards marine collision avoidance based on automatic route exchange DOI : 10.3182/20130918-4-JP-3022.00049
-
2013
Decentralised collision avoidance in a semi-collaborative multi-agent system DOI : 10.1007/978-3-642-40776-5_36
-
2012
Supervised task performance of an autonomous UAV swarm, supporting and implementing fire-fighting procedures
Teaching
Artificial Intelligence
Theory of statistical learning; types of applications: classification, regression, prediction, categorization, … neural networks (multilayer, RBF, …) ; kernel methods and Support Vector Machines (SVM); boosting; probabilistic graphical models (Bayesian networks); unsupervised learning for categorization (k-means, Kohonen topological maps, etc.); evolutionary algorithms and other meta-heuristics.
PhD supervision
- 2024 Motion prediction involving agent-to-agent interactions and multimodal modeling AZEVEDO TONÉ Caio
- 2022 Multimodal reasoning for geometric and acoustic scene reconstruction BRUNETTO Amandine
- 2021 Integrate expert knowledge into deep reinforcement learning methods for autonomous driving. CHEKROUN Raphaël