中文注视行为标注工具包(GBAT): 基于AI的自动标注工具，用于儿童与照护者互动的第一人称眼动追踪和视频数据

本文介绍基于深度学习的GazeBehavior Annotation Toolkit，用于简化儿童-看护者互动视频的数据预处理与特征提取，实现多视频同步与半自动标注，助力实时注意力、动作和语言交互研究，减少人工成本。

arXiv:2605.22962v1 Announce Type: new Abstract: Video recordings of child-caregiver interactions enable investigation of attentional dynamics during naturalistic behavior. Such multimodal recording also allows researchers to examine how attention interacts with action and language use in real time. However, manual annotation of such data is time-consuming. Here, we introduce GazeBehavior Annotation Toolkit, a deep-learning-based toolkit designed to facilitate three key processes in data preprocessing and feature extraction: post-hoc synchronization across multiple videos, semi-automatic annota