Scholarlink

<p>Online shopping platforms are experiencing rapid growth, necessitating effective product recommendation systems to enhance customer satisfaction by recommending visually similar products. Traditional statistical techniques often result in less accurate recommendations. This study encompasses two primary tasks: Query Product Classification and Product Recommendation Retrieval. The initial job employs a Deep Neural Network, which takes high-level feature representations derived from the Xception and VGG16 for the query image. The features are further analyzed to forecast the category of the product. The second step is retrieving analogous products by calculating cosine similarity, facilitating the identification of visually comparable items within the product database. By leveraging diverse feature representations, the proposed approach improves the precision and relevance of recommendations. Experimental evaluations on fashion product images and Shoe datasets demonstrate significant performance improvements over existing models, achieving accuracies of 91.76%, and 82.23% respectively. These results underscore the capability of the system to provide superior recommendations in online fashion shopping scenarios, emphasizing its effectiveness in improving customer experience.</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

A Two-Stage Deep Learning Approach for Optimizing Fashion Product Recommendations

  • 期刊论文

  • 发表日期：

    2025年4月3日

查看全文
  • Suvarna Buradagunta,
  • Sivadi Balakrishna

摘要

Online shopping platforms are experiencing rapid growth, necessitating effective product recommendation systems to enhance customer satisfaction by recommending visually similar products. Traditional statistical techniques often result in less accurate recommendations. This study encompasses two primary tasks: Query Product Classification and Product Recommendation Retrieval. The initial job employs a Deep Neural Network, which takes high-level feature representations derived from the Xception and VGG16 for the query image. The features are further analyzed to forecast the category of the product. The second step is retrieving analogous products by calculating cosine similarity, facilitating the identification of visually comparable items within the product database. By leveraging diverse feature representations, the proposed approach improves the precision and relevance of recommendations. Experimental evaluations on fashion product images and Shoe datasets demonstrate significant performance improvements over existing models, achieving accuracies of 91.76%, and 82.23% respectively. These results underscore the capability of the system to provide superior recommendations in online fashion shopping scenarios, emphasizing its effectiveness in improving customer experience.