Robust and Precise Sidewalk Detection with Ensemble Learning: Enhancing Road Safety and Facilitating Curb Space Management

The study highlights a robust ensemble model's effectiveness for accurate sidewalk detection, vital for both road safety and efficient curb space management. To evaluate the proposed ensemble model, three distinct datasets were utilized: Cityscapes, Ade20k, and the Boston Dataset. The results demonstrated the superiority of the ensemble model over its individual components, as manifested by mIOU scores of 93.1%, 90.3%, and 90.6% on the Cityscapes, Ade20k, and Boston datasets respectively, in optimal conditions. Under exposure to various noise types, such as Salt-and-Pepper and Speckle, across low to high intensities, the model revealed a steady, controlled decline in performance. This stands in contrast to the rapid drop experienced by individual models. Overall, the model demonstrated robustness and dependability. The ensemble model's resilience and dependability make a strong case for its application in enhancing road safety through reliable sidewalk detection and efficient curb space management.