Assessing the Performance of Instance Segmentation Algorithms for Indian Roads

This study assesses the effectiveness of the instance segmentation algorithm using YOLOv8 and its derivatives, in instance segmentation within Indian road environments. It aims to evaluate their ability to accurately detect and segment key elements such as roads, pedestrians/humans, and vehicles in the diverse roads of India. Performance metrics such as precision, recall, mAP50, and mAP50-95 will be utilized to compare different YOLOv8 variants. Additionally, the research will investigate the computational efficiency of these algorithms to gauge their suitability for potential real-time application in autonomous vehicles navigating Indian roads. By addressing the unique challenges and complexities presented by Indian road scenes, this study aims to contribute to the advancement of autonomous vehicle technology in challenging environments.