"BDRoadScenes: A Real-World Dataset for Multi-Class Vehicle Detection"

"This study has presented a real-world vehicle image dataset collected from busy streets in Bangladesh with around 500 mobile captured images that include different types of vehicles such as rickshaws cars vans ambulances microbuses trucks and buses. Many images have contained multiple vehicles in a single scene which has supported multi object detection in crowded traffic conditions. The dataset has undergone basic preprocessing where the images have been shuffled to reduce bias and human faces have been blurred to protect privacy. Although the image quality has remained relatively low the dataset has still captured diverse and realistic traffic situations which has made it a useful resource for developing and evaluating vehicle detection models in complex urban environments. "