RoadAnomaly

The 'RoadAnomaly' dataset is a benchmark used to evaluate the capability of models to detect out-of-distribution regions in semantic segmentation tasks, particularly in the context of safety-critical applications like autonomous driving.

Papers using RoadAnomaly (2)

Road-Aware Anomaly Segmentation with Query-Guided Polygons and CLIP in Autonomous Driving2026Training-Free Out-Of-Distribution Segmentation With Foundation Models2025