eduzhai > Applied Sciences > Engineering >

A Federated Approach for Fine-Grained Classification of Fashion Apparel

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: As online retail services proliferate and are pervasive in modern lives,applications for classifying fashion apparel features from image data arebecoming more indispensable. Online retailers, from leading companies tostart-ups, can leverage such applications in order to increase profit marginand enhance the consumer experience. Many notable schemes have been proposed toclassify fashion items, however, the majority of which focused upon classifyingbasic-level categories, such as T-shirts, pants, skirts, shoes, bags, and soforth. In contrast to most prior efforts, this paper aims to enable an in-depthclassification of fashion item attributes within the same category. Beginningwith a single dress, we seek to classify the type of dress hem, the hem length,and the sleeve length. The proposed scheme is comprised of three major stages:(a) localization of a target item from an input image using semanticsegmentation, (b) detection of human key points (e.g., point of shoulder) usinga pre-trained CNN and a bounding box, and (c) three phases to classify theattributes using a combination of algorithmic approaches and deep neuralnetworks. The experimental results demonstrate that the proposed scheme ishighly effective, with all categories having average precision of above 93.02 ,and outperforms existing Convolutional Neural Networks (CNNs)-based schemes.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×