CorVS Dataset: Visual Tracking Trajectories and Wearable Sensor Measurements in a Logistics Warehouse

Overview This dataset aims to develop and evaluate systems that identify individuals in videos by combining wearable sensors under real-world settings. It contains one hour of concurrent visual tracking trajectories and sensor measurements in a logistics warehouse. The trajectories were made using video footage from 19 cameras mounted on the ceiling. The sensor measurements were collected via smartphones carried by 28 workers. Sensor measurements and corresponding trajectories are labeled with common worker IDs. The associated paper presents the procedures for data collection and annotation. Please cite the paper when using these resources in your project. K. Kano, Y. Mori, S. Katayama, K. Urano, T. Yonezawa, and N. Kawaguchi, "CorVS+: Correspondence-Driven Association of Video Trajectories and Sensors for Identity-Aware Person Localization in Warehouses," 2026. https://arxiv.org/abs/2510.26369 Contents Trajectories 28 individuals with sensor + N individuals without sensor 463 tracks, 25 hours, 38 km in total 2.5 Hz Sensor measurements 28 individuals 27 hours in total 100 Hz Videos Pre-trained model for correspondence estimation Code The instructions and sample code for data processing and model inference are available at the associated repository.https://github.com/kazumakano/corvs-plus Contact Kazuma Kano Graduate School of Engineering, Nagoya University Email: [e-mail]