Scholarlink

<p>In this paper, we propose an Enhanced Content-Based Image Retrieval (CBIR) system designed for secure and personalized cloth recommendation, utilizing a hybrid algorithm integrated with Explainable AI (XAI) techniques and Region Adjacency Graph (RAG) descriptors. With the rapid growth of online shopping and fashion recommendation platforms, providing accurate, secure, and transparent recommendations has become essential. Traditional CBIR systems often face limitations in capturing complex clothing features and providing users with interpretable suggestions. To address these challenges, our approach combines multiple advanced techniques within a hybrid algorithm that enhances the feature extraction process. This includes leveraging RAG descriptors to model spatial relationships between regions in an image, which allows for more precise retrieval of similar clothing items based on shape, texture, and pattern features. Moreover, by incorporating Explainable AI into the recommendation framework, our system ensures that users are presented with clear, understandable justifications for each recommendation. This transparency improves user trust and offers insights into the decision-making process behind the algorithm’s choices. The secure aspect of our system is achieved by embedding security mechanisms to safeguard sensitive user data, ensuring that the recommendation process is not only accurate but also privacy-preserving. To validate the effectiveness of the proposed system, we conducted extensive experiments using diverse fashion datasets. The results demonstrate significant improvements in both retrieval accuracy and interpretability compared to traditional CBIR systems. Our system excels in identifying complex clothing attributes and providing relevant recommendations, while the XAI integration enhances user engagement by offering explanations for the suggested items. This research contributes to the evolving field of personalized fashion recommendation by proposing a secure, efficient, and explainable CBIR system capable of meeting modern user demands. Unlike previous CBIR-based fashion recommendation systems, which emphasize either accuracy or personalization, our work uniquely integrates three dimensions of explainability, region adjacency graph (RAG) descriptors, and secure data handling within a hybrid deep learning framework. This holistic approach advances personalized fashion recommendation by ensuring transparency, robustness, and privacy, making it a novel contribution in the e-commerce domain.</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

Enhanced CBIR based secure cloth recommendation system using hybrid algorithm with explainable AI and RAG descriptors

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月3日

查看全文
  • Akanksha Mrinali,
  • Pankaj Gupta

摘要

In this paper, we propose an Enhanced Content-Based Image Retrieval (CBIR) system designed for secure and personalized cloth recommendation, utilizing a hybrid algorithm integrated with Explainable AI (XAI) techniques and Region Adjacency Graph (RAG) descriptors. With the rapid growth of online shopping and fashion recommendation platforms, providing accurate, secure, and transparent recommendations has become essential. Traditional CBIR systems often face limitations in capturing complex clothing features and providing users with interpretable suggestions. To address these challenges, our approach combines multiple advanced techniques within a hybrid algorithm that enhances the feature extraction process. This includes leveraging RAG descriptors to model spatial relationships between regions in an image, which allows for more precise retrieval of similar clothing items based on shape, texture, and pattern features. Moreover, by incorporating Explainable AI into the recommendation framework, our system ensures that users are presented with clear, understandable justifications for each recommendation. This transparency improves user trust and offers insights into the decision-making process behind the algorithm’s choices. The secure aspect of our system is achieved by embedding security mechanisms to safeguard sensitive user data, ensuring that the recommendation process is not only accurate but also privacy-preserving. To validate the effectiveness of the proposed system, we conducted extensive experiments using diverse fashion datasets. The results demonstrate significant improvements in both retrieval accuracy and interpretability compared to traditional CBIR systems. Our system excels in identifying complex clothing attributes and providing relevant recommendations, while the XAI integration enhances user engagement by offering explanations for the suggested items. This research contributes to the evolving field of personalized fashion recommendation by proposing a secure, efficient, and explainable CBIR system capable of meeting modern user demands. Unlike previous CBIR-based fashion recommendation systems, which emphasize either accuracy or personalization, our work uniquely integrates three dimensions of explainability, region adjacency graph (RAG) descriptors, and secure data handling within a hybrid deep learning framework. This holistic approach advances personalized fashion recommendation by ensuring transparency, robustness, and privacy, making it a novel contribution in the e-commerce domain.