Short-Range Vehicle Door Opening Behavior Image Dataset

Description: This dataset contains high-quality synthetic images of vehicle door opening states captured within a short range of 0.5–2 meters, using both fisheye and pinhole cameras mounted around the vehicle body and door sides. It covers multiple vehicle types such as sedans, SUVs, MPVs, and pickup trucks, with varied door opening degrees including fully open, half-open, slightly open, and closed. Door positions include driver’s side, passenger’s side, and rear doors, presented in different combinations to reflect realistic opening scenarios. The images are high resolution (1920×1080) and simulate real-world sentinel mode perception setups.

The dataset is designed for vehicle door behavior recognition, collision warning systems, sentinel mode object perception, and visual model training for perimeter awareness in autonomous driving and security systems. It enables robust detection and classification of door positions and opening angles to enhance risk assessment and safety measures in various operational contexts.

Data was generated under simulated outdoor natural environments with diverse lighting conditions, including daytime, overcast, and nighttime scenarios. Capture settings incorporate multiple vehicle viewpoints—frontal, side, and diagonal perspectives—combined with door state variations to ensure comprehensive coverage. Vehicles are depicted in stationary positions, allowing clear analysis of door-related features without motion blur.

All samples are annotated in JSON format, including door state classification, opening angle estimation, and vehicle type identification. Label categories cover door position (driver/passenger/rear), opening degree (fully open, half-open, slightly open, closed), and vehicle type (sedan, SUV, MPV, pickup truck). Balanced representation across these categories ensures strong model generalization and robustness in real-world applications.

By integrating photorealistic rendering, human annotation, and diverse environmental and positional configurations, this dataset provides a valuable resource for developing high-accuracy door behavior recognition systems, enhancing vehicle safety alerts, and supporting intelligent surveillance capabilities.

Keywords: door opening detection vehicle door sentinel mode short-range perception door state classification collision warning

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

Metrics: Door state classification accuracy, opening angle estimation precision

Entities: Sedans, SUVs, MPVs, pickup trucks, vehicle doors

Geographies: China (urban, residential, and mixed-use road environments)

Price Range (USD): 4000–20000

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