Robust Traffic Sign Recognition Against Camera Failures

Failures of the vehicle camera may compromise correct acquisition frames, that are subsequently used by autonomous driving tasks. A clear understanding behavior tasks under such failure conditions, together with strategies to avoid safety is jeopardized, indeed necessary. This study analyses and improve performance Traffic Sign Recognition (TSR) systems for road vehicles possible occurrence failures. Our experimental assessment relies on three public datasets, which commonly benchmarking TSR systems. We artificially inject 13 different types failures into datasets. Then, we exploit deep neural networks (DNNs) classify either a single frame traffic sign or sequence (i.e., sliding window) frames. show windows significantly improves robustness classifier against altered confirm our observations through explainable AI, allows why classifiers have in case