Scholarlink

With the advancement of technology and the development of artificial intelligence, research methods in fashion design are continuously evolving. This research aims to use object detection techniques to analyze the attire of university students, exploring and uncovering fashion trends. Firstly, we selected YOLOv8 as object detection models, then trained the model and used the training set as well as adjusted parameters to improve accuracy. Secondly, we applied annotation tools to label collected photos, marking different types of clothing (e.g., tops, pants, shoes). To further enhance the quality and variety of the dataset, we leveraged a pre-trained model based on the DeepFashion2 [10] dataset. By leveraging big data and machine learning technologies, this project will establish an automated system capable of analyzing university students’ fashion styles in real-time and with accuracy, providing forward-looking insights for fashion design. In this research, we have chosen to focus on one of the most produced fashion items globally - T-shirts. Mass manufactured knit tops such as T-shirts, polo shirts, and sleeveless tops are chosen as they are one of the most ubiquitous items shown in our preliminary testing, but more importantly, as they are some of the most resource intensive apparel items to produce. In our Dataset creation we create 14 new categories of T-shirt with annotation tool RoboFlow [13] and segmentation technique, which represent different shape, size and color patterns of T-shirts. The students of the Department of Fashion Design and management were the main contributors to this Dataset. Fashion design is not only an art but also a discipline closely related to society and culture. Traditional methods of analyzing fashion trends often rely on surveys and street observations, which are time-consuming and subjective. With the development of artificial intelligence and big data technologies, applying object detection techniques to attire analysis can enhance efficiency and provide more objective and comprehensive data.

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

Using Object Detection Methods to Detect Fashion Trends in University Students’ Attire

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wei-Her Hsieh,
  • Md. Rabius Sany Apu,
  • Jatin Thapa,
  • Owen Kumala

摘要

With the advancement of technology and the development of artificial intelligence, research methods in fashion design are continuously evolving. This research aims to use object detection techniques to analyze the attire of university students, exploring and uncovering fashion trends. Firstly, we selected YOLOv8 as object detection models, then trained the model and used the training set as well as adjusted parameters to improve accuracy. Secondly, we applied annotation tools to label collected photos, marking different types of clothing (e.g., tops, pants, shoes). To further enhance the quality and variety of the dataset, we leveraged a pre-trained model based on the DeepFashion2 [10] dataset. By leveraging big data and machine learning technologies, this project will establish an automated system capable of analyzing university students’ fashion styles in real-time and with accuracy, providing forward-looking insights for fashion design. In this research, we have chosen to focus on one of the most produced fashion items globally - T-shirts. Mass manufactured knit tops such as T-shirts, polo shirts, and sleeveless tops are chosen as they are one of the most ubiquitous items shown in our preliminary testing, but more importantly, as they are some of the most resource intensive apparel items to produce. In our Dataset creation we create 14 new categories of T-shirt with annotation tool RoboFlow [13] and segmentation technique, which represent different shape, size and color patterns of T-shirts. The students of the Department of Fashion Design and management were the main contributors to this Dataset. Fashion design is not only an art but also a discipline closely related to society and culture. Traditional methods of analyzing fashion trends often rely on surveys and street observations, which are time-consuming and subjective. With the development of artificial intelligence and big data technologies, applying object detection techniques to attire analysis can enhance efficiency and provide more objective and comprehensive data.