Physically interpretable probabilistic domain characterization
(2024) Computer Vision – ACCV 2024 Workshops
Files
Halin2024Physically.pdf
Open Access - Adobe PDF
- 25.7 MB
Details
- Authors
- et. al.
- Abstract
- Characterizing domains is essential for models analyzing dy- namic environments, as it allows them to adapt to evolving conditions or to hand the task over to backup systems when facing conditions out- side their operational domain. Existing solutions typically characterize a domain by solving a regression or classiőcation problem, which limits their applicability as they only provide a limited summarized description of the domain. In this paper, we present a novel approach to domain characterization by characterizing domains as probability distributions. Particularly, we develop a method to predict the likelihood of different weather conditions from images captured by vehicle-mounted cameras by estimating distributions of physical parameters using normalizing ŕows. To validate our proposed approach, we conduct experiments within the context of autonomous vehicles, focusing on predicting the distribution of weather parameters to characterize the operational domain. This do- main is characterized by physical parameters (absolute characterization) and arbitrarily predeőned domains (relative characterization). Finally, we evaluate whether a system can safely operate in a target domain by comparing it to multiple source domains where safety has already been established. This approach holds signiőcant potential, as accurate weather prediction and effective domain adaptation are crucial for au- tonomous systems to adjust to dynamic environmental conditions.
- Affiliations
Citations
Halin, A., Piérard, S., Vandeghen, R., Gerin, B., Zanella, M., & et al. (2024). Physically interpretable probabilistic domain characterization. Lecture Notes in Computer Science, 15482(1), 17-35. https://hdl.handle.net/2078.5/267822 (Original work published 2024)