Yu-Sian Lin, Chow-Sing Lin: Enhancing the accuracy of prototype learning in road anomaly segmentation by adding adversarial perturbations to data. Multim. Tools Appl. 83(18): 55627-55640 (2024)