[4F3-GS-3] Knowledge utilization and sharing:

[4F3-GS-3-04] A Construction of the Data Set for Classification of Used Clothes

〇Chie Mizunuma¹, Tomoko Ikeya¹, Hiromi Kurosaki¹, Norika Note¹, Noriyuki Okumura¹, Yasushi Kami², Miwako Hanada¹ (1. Kobe Shoin Women's University, 2. Hiroshima Institute of Technology)

[[Online]]

Keywords:data set, used clothes, recycle

The fashion industry poses a significant environmental burden from sourcing raw materials to disposal, making it an international challenge to address. According to a survey by the Ministry of the Environment, 95% of clothing disposed of as household waste is either incinerated or landfilled. This research aims to construct a system using AI to automatically determine whether unwanted clothing items can be reused or recycled, thereby reducing the amount of clothing sent to waste.In this report, we discuss the method for creating a dataset of clothing images necessary for machine learning, as there is no open dataset specifically targeting clothing items that fall within the borderlines of reuse or recycling. Images were obtained from clothing collected mainly from thrift stores dealing with low-cost items. These items were photographed from multiple directions using several cameras with varying resolutions, resulting in a total of 14,105 images. Furthermore, interviews were conducted with both sorters and individuals lacking sorting knowledge to gather their sorting processes and reasoning for reuse or recycling, which were then used to tag the collected images. Additionally, for social implementation, a system was considered to coordinate and present reusable clothing items together for potential outfits.