In-Vehicle Short-Range Multi-Vehicle Type Image Dataset

Description: This dataset provides a large-scale collection of short-range vehicle images captured within 0.5 to 2 meters using an in-vehicle pinhole camera system. It covers multiple vehicle types, colors, lighting conditions, and shooting angles, representing common sentinel mode application scenarios such as residential areas, industrial zones, commercial districts, and parking lots. The dataset includes approximately 100,000 high-resolution frames (1920×1080) with a total data volume of around 260 GB, offering extensive coverage of near-field perception requirements.

The dataset is designed for close-range vehicle detection and classification in autonomous driving and sentinel mode analysis. It supports applications including near-field object recognition, vehicle type classification, and behavioral monitoring for vehicles in static or low-speed states. By capturing diverse vehicle appearances across environments, it ensures strong generalization performance in real-world deployments.

All samples are annotated with vehicle bounding boxes and classification labels, provided in JSON format. Label categories include passenger car, SUV, MPV, pickup truck, light truck, heavy truck, van, motorcycle, bicycle, and electric tricycle, with balanced representation across classes to support high-accuracy detection and classification.

The capture process simulates varied environmental and lighting conditions such as daytime, nighttime, and rainy weather. Data includes multi-angle views (front, side, rear) and covers a variety of parking and traffic scenarios, ensuring robustness against variations in scene layout, background complexity, and object occlusion.

By combining high-resolution imagery, human annotation, and diverse environmental coverage, this dataset provides a reliable foundation for training and validating near-field vehicle recognition models, enhancing sentinel mode capabilities, and improving short-range autonomous perception systems.

Keywords: vehicle recognition sentinel mode short-range perception vehicle classification near-field detection

Sector: Intelligent Transportation / Autonomous Driving / Vehicle Recognition / Surveillance Systems

Metrics: Vehicle detection accuracy, classification precision per vehicle type Entities: Passenger cars, SUVs, MPVs, pickup trucks, light trucks, heavy trucks, vans, motorcycles, bicycles, electric tricycles

Geographies: China (residential, industrial, commercial, and parking lot environments)

Price Range (USD): 4000–20000

Source type: Synthetic data generation from simulation environment, human annotation