Multi-View In-Cabin Monitoring System for Public Transport Vehicles

Researchers introduce a multi-view in-cabin monitoring dataset for public transport vehicles, featuring synchronized RGB and depth images from four cameras and LiDAR data collected from a German city bus. The dataset includes 9,136 annotated samples with 3D pose estimates and bounding boxes, along with benchmarked detection models to advance multi-view perception systems for autonomous public transportation.

This dataset release represents a focused contribution to autonomous vehicle perception research, specifically addressing the underexplored domain of interior cabin monitoring rather than external environment sensing. The synchronized multi-modal data from RGB cameras, depth sensors, and rotating LiDAR provides researchers with the hardware diversity needed to develop robust occupant detection and tracking systems—critical for safety in automated public transport where passenger monitoring informs operational decisions.

The research fills a practical gap in autonomous vehicle development. While most public datasets emphasize road perception and obstacle detection, interior monitoring remains essential for Level 4 autonomous buses operating in urban environments. German cities' investment in digitalized public transport infrastructure positions this work within Europe's broader push toward intelligent mobility systems, where passenger safety and comfort monitoring directly impacts regulatory approval and public acceptance.

The inclusion of calibration pipelines, pseudo-labeling methodologies, and nuScenes-format conversion demonstrates the authors' focus on practical reproducibility rather than isolated benchmarking. By providing open-source tools and comparative evaluations of established models like Lift-Splat-Shoot and BEVFusion, the dataset enables smaller research groups to participate in cabin perception development without requiring extensive infrastructure or annotation resources.

For the autonomous vehicle industry, this work supports incremental progress toward fully automated public transit systems. The European emphasis on safety and standardization means interior monitoring systems will face regulatory scrutiny; datasets like this accelerate the development cycle. However, the modest sample size (9,136 scenes) limits direct commercial deployment, positioning this primarily as a research enabler rather than a production-ready solution.

• →Multi-view in-cabin monitoring dataset released with 9,136 synchronized RGB-depth-LiDAR samples from an automated German city bus
• →Includes calibration pipeline and pseudo-labeling methodology generating 3D pose estimates and oriented bounding boxes for occupants
• →Benchmarks multi-view 3D detection models and provides nuScenes-format conversion for standardized evaluation
• →Addresses underexplored interior monitoring domain critical for Level 4 autonomous public transport safety and passenger management
• →Open-source tools and comparative model evaluations enable broader research participation in cabin perception development