Shizuoka University
Okabe Lab

Video content analysis
using face identification models and Generative AI

Azumi Tsunoda

Shizuoka University

Makoto Okabe

Shizuoka University

研究のTeaser画像

Flow of the video analysis system (blue: input to the system, yellow: output to the user)
InsightFace identifies the characters and matches them with the video script to analyze the story and search for related scenes.

Abstract

Using face recognition by InsightFace and natural language processing by ChatGPT, we search for characters and analyze their stories in videos. By analyzing the script of a video with timecode, the system outputs spoiler-free information on the story and human relationships up to the specified scene. This system provides a rich viewing experience by supporting and facilitating a deeper understanding of the video content.

Paper

Bachelor Thesis

IPSJ SIG-CGVI, 197th conference

Video

BibTeX citation

        @Misc{tsunoda.azumi.21,
            Kauthor =    {角田 梓美},
            title =     {顔識別モデルと生成AIを用いた動画内容の解析},
            howpublished = {Bachelor Thesis},
            month =     2,
            year =      2025,
        }