Latent Discriminant deterministic Uncertainty
This work addresses the need for computationally efficient uncertainty estimation in real-world autonomous systems, offering an incremental improvement over existing deterministic methods for domain-specific computer vision tasks.
The paper tackles the challenge of scalable predictive uncertainty estimation for deep neural networks in autonomous driving perception by proposing a deterministic uncertainty method that relaxes the Lipschitz constraint and uses a discriminant latent space with prototypes, achieving competitive results compared to state-of-the-art Deep Ensembles on tasks like image classification, segmentation, and depth estimation.
Predictive uncertainty estimation is essential for deploying Deep Neural Networks in real-world autonomous systems. However, most successful approaches are computationally intensive. In this work, we attempt to address these challenges in the context of autonomous driving perception tasks. Recently proposed Deterministic Uncertainty Methods (DUM) can only partially meet such requirements as their scalability to complex computer vision tasks is not obvious. In this work we advance a scalable and effective DUM for high-resolution semantic segmentation, that relaxes the Lipschitz constraint typically hindering practicality of such architectures. We learn a discriminant latent space by leveraging a distinction maximization layer over an arbitrarily-sized set of trainable prototypes. Our approach achieves competitive results over Deep Ensembles, the state-of-the-art for uncertainty prediction, on image classification, segmentation and monocular depth estimation tasks. Our code is available at https://github.com/ENSTA-U2IS/LDU