Scholarlink

In recent years, the interior design industry has increasingly embraced AI and machine learning technologies to provide personalized design recommendations and streamline the design process. However, analyzing and clustering large sets of interior design images to extract meaningful patterns remains a challenge. This study addresses this problem by applying convolutional neural networks (CNN) for feature extraction combined with three different dimensionality reduction techniques—Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP)—to cluster and analyze interior design images. Using K-Means clustering, we evaluated the effectiveness of each dimensionality reduction technique through silhouette scores, identifying PCA as the most effective method with a score of 0.42 compared to 0.33 for t-SNE and 0.32 for UMAP. The analysis of clusters revealed distinct design trends, such as traditional, modern, minimalist, and luxury styles. The study’s implications highlight the potential for developing AI-driven interior design recommendation systems that offer personalized suggestions based on user preferences. However, the study also identifies limitations in the dataset and methodology, such as the lack of metadata integration and diverse design styles. Future work will explore deeper integration of metadata and more advanced AI models for enhanced recommendation accuracy.

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

Clustering and Feature Extraction for Interior Design Images Using Machine Learning and Dimensionality Reduction Techniques

  • 章节

  • 发表日期：

    2025年

查看全文
  • Said A. Salloum,
  • Ra’ed Masa’deh,
  • Hanan M. Taleb,
  • Khaled Shaalan

摘要

In recent years, the interior design industry has increasingly embraced AI and machine learning technologies to provide personalized design recommendations and streamline the design process. However, analyzing and clustering large sets of interior design images to extract meaningful patterns remains a challenge. This study addresses this problem by applying convolutional neural networks (CNN) for feature extraction combined with three different dimensionality reduction techniques—Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP)—to cluster and analyze interior design images. Using K-Means clustering, we evaluated the effectiveness of each dimensionality reduction technique through silhouette scores, identifying PCA as the most effective method with a score of 0.42 compared to 0.33 for t-SNE and 0.32 for UMAP. The analysis of clusters revealed distinct design trends, such as traditional, modern, minimalist, and luxury styles. The study’s implications highlight the potential for developing AI-driven interior design recommendation systems that offer personalized suggestions based on user preferences. However, the study also identifies limitations in the dataset and methodology, such as the lack of metadata integration and diverse design styles. Future work will explore deeper integration of metadata and more advanced AI models for enhanced recommendation accuracy.