Visual Categorization of Apparel Products Combining Bag of Visual Words (BoVW) and Color and Edge Directivity Descriptor (CEDD)

• Joyce Miranda dos Santos UFAM
• Josiane R. da Silva UFAM
• Edleno Silva de Moura UFAM
• Caio Moura Daoud UFAM

Resumo

In e-commerce scenario, identify in advance the category of a product is fundamental to improving the quality of the returned ranking for a query. In this article, we propose an alternative to the task of categorization of apparel products which explores the combination of representation generated by Bag of Visual Words (BoVW) method with features extracted by a traditional content-based image descriptor, the Color and Edge Directivity Descriptor (CEDD). Our experiments show that the proposed combination strategy is viable, competitive and statistically higher compared to baselines