著者

棚本 悠太 綿貫 啓一 楓 和憲 村松 慶一

出版者

一般社団法人 日本機械学会

雑誌

年次大会

巻号頁・発行日

vol.2019, 2019

---

<p>The purpose of this study is to estimate the condition of second-hand goods from images using machine learning methods. Assuming that the condition evaluation of second-hand goods has a relationship with the damage that exists in them, we proposed and verified a method to estimate the condition of second-hand goods based on information of the damage. In this study, we chose men's dress shoes for condition evaluation and chose four types of damage: wrinkles, worn out of heel, dirt on outsole, dirt on inside as damage considered to be related to condition evaluation. (1)We create a CNN model that determines the presence or absence of damage, and (2) create a classification tree model that evaluates the state of the product from information on the presence or absence of damage and (3)We evaluated the condition of second-hand goods by combining two created models. As a result of verification, it was shown that the proposed method could evaluate the condition of used goods.</p>