Scholarlink

<p>The fashion industry is experiencing a significant digital transformation driven by Deep Learning (DL) and the Internet of Things (IoT), which facilitates personalized and sustainable apparel design. Smart devices and wearable sensors collect extensive user data for tailored clothing recommendations. Traditional design methods, relying on manual measurements and size charts, often lead to poor fit and user dissatisfaction, highlighting the necessity for intelligent systems that can adapt to individual preferences and characteristics. This research proposes a novel DL-based model, the Secretary Bird-driven Intelligent Long Short-Term Memory network integrated with the Weighted Generative Adversarial Network (SB-Int-LSTM-WGANet), designed to optimize personalized clothing design, classify individual style preferences, and generate realistic clothing images. Data were collected from the Personalized Clothing &amp; Body Measurements Dataset, including 3D body measurements, posture dynamics, and lifestyle activity information. Preprocessing included image resizing, Gaussian filtering, and Z-score normalization to ensure consistency and remove noise. Convoloutional Neural Network (CNN)-based feature extraction captured fine-grained visual and dimensional characteristics, followed by feature-level fusion to combine multi-source inputs effectively. The Int-LSTM module captures temporal and sequential patterns in user behaviour for short-term and long-term style preference classification. The SB optimization module enhances adaptability to individual characteristics, while the WGANet module generates high-quality, realistic clothing designs and enables interactive virtual try-on simulations. Experimental results demonstrate that SB-Int-LSTM-WGANet achieves Fréchet Inception Distance (FID) (1.024), Inception Score (IS) (6.093), and Structural Similarity Index Measure (SSIM) (0.921), outperforming traditional methods. In conclusion, the research presents a robust, data-driven framework for sustainable and user-centric clothing design, effectively bridging personalization, efficiency, and practicality in the modern fashion industry.</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

Development of an intelligent system for personalized clothing design based on deep learning and the internet of things

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月3日

查看全文
  • Xiaoyang Liu,
  • He Zhao

摘要

The fashion industry is experiencing a significant digital transformation driven by Deep Learning (DL) and the Internet of Things (IoT), which facilitates personalized and sustainable apparel design. Smart devices and wearable sensors collect extensive user data for tailored clothing recommendations. Traditional design methods, relying on manual measurements and size charts, often lead to poor fit and user dissatisfaction, highlighting the necessity for intelligent systems that can adapt to individual preferences and characteristics. This research proposes a novel DL-based model, the Secretary Bird-driven Intelligent Long Short-Term Memory network integrated with the Weighted Generative Adversarial Network (SB-Int-LSTM-WGANet), designed to optimize personalized clothing design, classify individual style preferences, and generate realistic clothing images. Data were collected from the Personalized Clothing & Body Measurements Dataset, including 3D body measurements, posture dynamics, and lifestyle activity information. Preprocessing included image resizing, Gaussian filtering, and Z-score normalization to ensure consistency and remove noise. Convoloutional Neural Network (CNN)-based feature extraction captured fine-grained visual and dimensional characteristics, followed by feature-level fusion to combine multi-source inputs effectively. The Int-LSTM module captures temporal and sequential patterns in user behaviour for short-term and long-term style preference classification. The SB optimization module enhances adaptability to individual characteristics, while the WGANet module generates high-quality, realistic clothing designs and enables interactive virtual try-on simulations. Experimental results demonstrate that SB-Int-LSTM-WGANet achieves Fréchet Inception Distance (FID) (1.024), Inception Score (IS) (6.093), and Structural Similarity Index Measure (SSIM) (0.921), outperforming traditional methods. In conclusion, the research presents a robust, data-driven framework for sustainable and user-centric clothing design, effectively bridging personalization, efficiency, and practicality in the modern fashion industry.