Researchers have introduced BMD-45, a new large-scale dataset designed to improve vehicle detection in urban traffic environments found in developing cities. This dataset contains over 45,000 images with 480,000 bounding boxes, featuring 14 specific vehicle categories including region-specific ones like auto-rickshaws. Current models trained on existing benchmarks show a significant performance drop when applied to the diverse and dense traffic conditions present in BMD-45, highlighting the need for more geographically varied training data. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Provides a benchmark for developing more robust perception systems for underrepresented urban traffic environments.
RANK_REASON Release of a new academic dataset for computer vision research.