Analysis of Bus Stop Infrastructure Using Deep Learning-Based Object Detection: A Case of Thrissur, Kerala, India

Bus transport forms the most critical and widely used component of the public road transportation system in India. Bus stops are therefore the most critical nodes connecting passengers to this system. Using a deep learning-based object detection model, this research provides a detailed analysis of the bus stop infrastructure in Thrissur, Kerala, India. The model is applied to a custom dataset of 400 field and Google Street View (GSV) images, and a comprehensive audit is conducted of 80 stops, benchmarking them against Indian road development standards. The findings reveal a large difference in time cost between the methods and demonstrate a highly reliable deep learning-based object detection system. In parallel, a `de facto' barrier is identified: a clear spatial inequality between the urban core and the periphery systematically excludes accessibility-dependent groups from safe and equitable access to public transport. Crucially, a `Modern but Inaccessible' paradox was observed: 100% of surveyed bus stops had roof shelters, yet only 2.5% complied with fundamental accessibility standards, a critical failure seen even in the newest structures. This study concludes that deep learning models such as YOLOv8 are among the most effective (recall of 75.3%, precision of 71.7%) and time-efficient (about 15 seconds to survey 80 bus stops after initial development) methods for large-scale urban infrastructural analysis.