Auto-RickshawImageBD: An Image Dataset for Auto-Rickshaw Detection

The Auto-RickshawImageBD is designed for detecting and recognizing auto-rickshaws in complex urban traffic scenes using deep learning models. The dataset comprises 1,331 high-resolution images collected from various urban and suburban locations in Dhaka, Narayanganj, Gazipur, Savar, Cumilla, Narsindi and Khulna cities of Bangladesh.
Images were captured using smartphones and portable cameras at different times of the day to ensure diversity in lighting conditions. Both daytime and nighttime images were included to simulate real-world traffic scenarios and enhance model robustness under varying illumination.

All of the 1,331 images are annotated using Label Studio following the YOLO object detection format, labeling auto-rickshaw instances (Class 1) and non-auto rickshaw vehicles (Class 2). Each bounding box was manually verified to ensure high annotation quality.

The Auto-RickshawImageBD dataset is well-suited for tasks such as object detection, vehicle classification, and traffic monitoring in developing urban contexts where auto-rickshaws are prevalent.

Please cite our research paper based on this dataset if you use the dataset.