Scholarlink

The paper presents the research agenda of the Socially-Aware Learning through Interactions in Crowded Environments (SALICE) project. The ambition of the project is to build predictive models of human behaviors in crowded scenes, in order to design and implement socially-aware multi-agent systems. The project is an innovative attempt to model interactional phenomena through the combination of psychological and engineering concepts. To this end, the current paper presents the methodology adopted to create an appropriately annotated dataset, that can be used for supervised learning, and serve as a benchmark for assessing the efficiency of autonomous systems in forecasting human trajectories in crowded environments. The proposed methodology, and the related annotation ontology developed within this context, provide a replicable annotation procedure that could be adopted by future works developing forecasting and/or human motion detection models.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Annotated Datasets for Trajectories’ Prediction: A Research Agenda

  • 章节

  • 发表日期：

    2025年

查看全文
  • Claudia Greco,
  • Giovanni Di Gennaro,
  • Marialucia Cuciniello,
  • Terry Amorese,
  • Maria Santina Ler,
  • Gennaro Cordasco,
  • Amedeo Buonanno,
  • Francesco A. N. Palmieri,
  • Anna Esposito

摘要

The paper presents the research agenda of the Socially-Aware Learning through Interactions in Crowded Environments (SALICE) project. The ambition of the project is to build predictive models of human behaviors in crowded scenes, in order to design and implement socially-aware multi-agent systems. The project is an innovative attempt to model interactional phenomena through the combination of psychological and engineering concepts. To this end, the current paper presents the methodology adopted to create an appropriately annotated dataset, that can be used for supervised learning, and serve as a benchmark for assessing the efficiency of autonomous systems in forecasting human trajectories in crowded environments. The proposed methodology, and the related annotation ontology developed within this context, provide a replicable annotation procedure that could be adopted by future works developing forecasting and/or human motion detection models.