This article discusses a new system for pedestrian data collection that is designed to aid in research on social navigation and pedestrian-robot interactions. The system uses machine learning methods and is equipped with a semi-autonomous labeling pipeline. It also features a label correction web app for human verification of automated pedestrian tracking outcomes. The system is capable of collecting large-scale data in diverse environments, and its resulting dataset, the TBD Pedestrian Dataset, is larger and richer in information compared to previous datasets. This new system and dataset can support new research opportunities in the field.
Publication date: 2 Oct 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2309.17187
