Bangladesh Urban Traffic Dataset

This production-grade urban traffic dataset captures complex, real-world interactions from Dhaka, Bangladesh, providing invaluable data for evaluating and training perception systems. Key features include: • Diverse mixed traffic including rickshaws, motorcycles, cars, buses, and trucks • Detailed capture of pedestrian behavior and market congestion • Urban intersections and informal road scenarios • GPS, IMU, route, time, weather, and scene-level metadata • Coverage for low-light conditions and dense interactions Designed to address limitations of simulated or structured datasets, this collection provides exposure to high-entropy environments where traditional assumptions often fail. It emphasizes informal traffic systems with dense interactions, lane-less movement, and implicit human negotiation, which are crucial for developing robust perception and navigation capabilities. Teams can isolate high-risk, high-uncertainty scenarios using segment-level metadata for targeted training and evaluation. Ideal for engineering teams developing systems for autonomous navigation, advanced driver-assistance, and robotics that must perform reliably beyond controlled environments. Researchers in computer vision and mobile robotics will find this data highly beneficial for assessing model performance in challenging, unstructured settings.