Scholarlink

<p>The objective of important people identification is to recognize the individuals who have the most significant impact in a social setting. However, current models exhibit subpar performance when confronted with a substantial volume of individuals, notable disparities in the significance of individuals, and intricate interplays across individuals. This research proposes an important people detection approach utilizing the Multi-Level Key Actor Interaction Graph (MLKAIG) to address these issues. The objective is to identify a key actor in a scenario by reducing the density of the actor graph and use the Multi-Level Key Actor Interaction Graph Attention Network to capture the complex interactions among individuals. In addition, MLKAIG incorporates a multi-information coding module that combines the significance of various information during the detection phase to enhance the model’s performance. MLKAIG considers the influence of its own feature information on the variability of the dataset. Validation of the proposed model’s effectiveness is confirmed through ablation experiments using several modules and hyperparameters. The model demonstrates superior performance compared to previous research when evaluated on the MS and NCAA datasets.</p>

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

Important people detection based on multi-level key actor interaction graph

  • 期刊论文

  • 发表日期：

    2025年3月10日

查看全文
  • Haifeng Sang,
  • Yanming Guo

摘要

The objective of important people identification is to recognize the individuals who have the most significant impact in a social setting. However, current models exhibit subpar performance when confronted with a substantial volume of individuals, notable disparities in the significance of individuals, and intricate interplays across individuals. This research proposes an important people detection approach utilizing the Multi-Level Key Actor Interaction Graph (MLKAIG) to address these issues. The objective is to identify a key actor in a scenario by reducing the density of the actor graph and use the Multi-Level Key Actor Interaction Graph Attention Network to capture the complex interactions among individuals. In addition, MLKAIG incorporates a multi-information coding module that combines the significance of various information during the detection phase to enhance the model’s performance. MLKAIG considers the influence of its own feature information on the variability of the dataset. Validation of the proposed model’s effectiveness is confirmed through ablation experiments using several modules and hyperparameters. The model demonstrates superior performance compared to previous research when evaluated on the MS and NCAA datasets.