MY-VID: A Malaysian Vehicle Image Dataset for Intelligent Transportation System and Road Safety

MY-VID v2 is the enhanced second release of the Malaysian Vehicle Image Dataset, featuring high-quality traffic images captured from a primary Malaysian expressway environment. While MY-VID v1 contained 8,832 images with only two broad categories, MY-VID v2 upgrades the dataset by adopting the six-class taxonomy established by the Malaysian Public Works Department (Jabatan Kerja Raya or JKR). This granular classification offers a precise representation of Malaysian traffic composition, distinguishing between specific vehicle types such as heavy lorries, light vans, and motorcycles. Although focused on a targeted highway domain, the dataset captures a rich variety of temporal and environmental conditions, including day and night scenes, diverse weather patterns (rain, fog, sunny), and fluctuating traffic densities. This focused operational domain allows for robust testing of AI models against the specific challenges of Malaysian expressway traffic, addressing the generalization gap often found in Western-centric datasets (e.g., COCO) when applied to Southeast Asian vehicle models (e.g., Proton, Perodua). Features Decsription Total Images 8,832 images Total Vehicle Annotations 47,975 vehicles Image Format JPG Annotation Format YOLO (.txt format) Classes Class 1 (Cars and Taxis) Class 2 (Vans and Utilities) Class 3 (Medium Lorries) Class 4 (Heavy Lorries) Class 5 (Buses) Class 6 (Motorcycles) Capture Conditions Different times of day (daytime to nighttime) Various weather conditions (sunny, cloudy, rainy) Across different days (weeksdays, weekends, public holidays) Varying traffic densitives (light, moderate and heavy traffic)