Scholarlink

<p>Social networks have become one of the main structures for representing people’s relationships, so building a structure that accurately describes relationships is an essential issue for the scientific community focused on analyzing relationships. Currently, the construction of social networks is a research problem due to the complexity of measuring how connected people are to each other. In our work, we propose a social network construction approach to represent social relationships based on facial analysis from a collection of images or video frames, called FA-SNet. Instead of measuring the connectedness strength of relationships based on the frequency of joint appearances present in a collection, we measure how connected people are; that is, we provide a connectivity score that represents the strength or closeness of relationships. This connectivity information is obtained from a connectivity matrix calculated by measuring factors based on facial analysis, such as ‘co-occurrence’, ‘closeness’, ‘connection’, ‘empathy’, and ‘happiness’. The social network is composed of nodes interconnected by edges, where the nodes represent the individuals and the edges express their relationships. The key idea of the obtained graph is that the closer are the nodes, the higher the connectivity between them. We evaluated the proposed approach through four experiments using different datasets, such as two summary clips from the television series <i>Pride and Prejudice</i>, a video from the film <i>Les Misérables</i>, a summary clip from the film <i>Emma</i>, and a set of images from a private wedding celebration. The effectiveness of the constructed networks was measured using the Graph Edit Distance metric, which shows how similar the proposed network is to reference networks. The results obtained from the proposed approach show potential in identifying relationships between individuals and quantifying the strength of these relationships using connectivity information. We believe our methodology will be a contribution for future studies in relationship analysis. The project page can be found in: <a href="https://github.com/Toscanojasiel/FA-SNet_Social_Network_Construction">https://github.com/Toscanojasiel/FA-SNet_Social_Network_Construction</a></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

FA-SNet: social network construction to represent social relationships based on facial analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月19日

查看全文
  • Jasiel Toscano,
  • Florencia Valdez,
  • Domingo Mery

摘要

Social networks have become one of the main structures for representing people’s relationships, so building a structure that accurately describes relationships is an essential issue for the scientific community focused on analyzing relationships. Currently, the construction of social networks is a research problem due to the complexity of measuring how connected people are to each other. In our work, we propose a social network construction approach to represent social relationships based on facial analysis from a collection of images or video frames, called FA-SNet. Instead of measuring the connectedness strength of relationships based on the frequency of joint appearances present in a collection, we measure how connected people are; that is, we provide a connectivity score that represents the strength or closeness of relationships. This connectivity information is obtained from a connectivity matrix calculated by measuring factors based on facial analysis, such as ‘co-occurrence’, ‘closeness’, ‘connection’, ‘empathy’, and ‘happiness’. The social network is composed of nodes interconnected by edges, where the nodes represent the individuals and the edges express their relationships. The key idea of the obtained graph is that the closer are the nodes, the higher the connectivity between them. We evaluated the proposed approach through four experiments using different datasets, such as two summary clips from the television series Pride and Prejudice, a video from the film Les Misérables, a summary clip from the film Emma, and a set of images from a private wedding celebration. The effectiveness of the constructed networks was measured using the Graph Edit Distance metric, which shows how similar the proposed network is to reference networks. The results obtained from the proposed approach show potential in identifying relationships between individuals and quantifying the strength of these relationships using connectivity information. We believe our methodology will be a contribution for future studies in relationship analysis. The project page can be found in: https://github.com/Toscanojasiel/FA-SNet_Social_Network_Construction