AI summaryⓘ
The authors studied how autonomous driving systems can detect unusual or new objects that they weren't trained on, which is important for safety. They focused on Visual Anomaly Detection (VAD), a technique that highlights unfamiliar parts of an image to alert the driver without knowing what the hazard looks like in advance. They tested eight VAD methods on a large synthetic dataset called AnoVox, using different neural network designs from big to small. Their findings show VAD works well for road scenes, and a lightweight model called Tiny-Dinomaly offers a good balance between accuracy and efficiency, making it suitable for real-world use. This work helps make self-driving cars safer by improving how they recognize unexpected obstacles.
Autonomous drivingMachine visionVisual Anomaly DetectionPixel-level anomaly mapsDeep learning backbonesEdge deploymentSynthetic datasetsTiny-DinomalyAnoVoxLocalization performance
Authors
Fabrizio Genilotti, Arianna Stropeni, Gionata Grotto, Francesco Borsatti, Manuel Barusco, Davide Dalle Pezze, Gian Antonio Susto
Abstract
The reliability of a machine vision system for autonomous driving depends heavily on its training data distribution. When a vehicle encounters significantly different conditions, such as atypical obstacles, its perceptual capabilities can degrade substantially. Unlike many domains where errors carry limited consequences, failures in autonomous driving translate directly into physical risk for passengers, pedestrians, and other road users. To address this challenge, we explore Visual Anomaly Detection (VAD) as a solution. VAD enables the identification of anomalous objects not present during training, allowing the system to alert the driver when an unfamiliar situation is detected. Crucially, VAD models produce pixel-level anomaly maps that can guide driver attention to specific regions of concern without requiring any prior assumptions about the nature or form of the hazard. We benchmark eight state-of-the-art VAD methods on AnoVox, the largest synthetic dataset for anomaly detection in autonomous driving. In particular, we evaluate performance across four backbone architectures spanning from large networks to lightweight ones such as MobileNet and DeiT-Tiny. Our results demonstrate that VAD transfers effectively to road scenes. Notably, Tiny-Dinomaly achieves the best accuracy-efficiency trade-off for edge deployment, matching full-scale localization performance at a fraction of the memory cost. This study represents a concrete step toward safer, more responsible deployment of autonomous vehicles, ultimately improving protection for passengers, pedestrians, and all road users.