There are many visually impaired people who use cosmetics, but it is said that they feel uncomfortable in distinguishing products with similar shapes. Therefore, if there is a system recognizing the kind and brand of cosmetics with similar shape (such...
There are many visually impaired people who use cosmetics, but it is said that they feel uncomfortable in distinguishing products with similar shapes. Therefore, if there is a system recognizing the kind and brand of cosmetics with similar shape (such as lip, mascara, eyebrow, and eyeliner) and informs them by voice, it will be able to relieve the inconvenience of the visually impaired. This paper proposes a system that automatically recognizes the type and brand of cosmetics from images taken with a mobile phone. For four types of cosmetics, we constructed a data set by directly taking 4400 cosmetics photos of two brands. As a result of training Yolov3 with this dataset, about 89.7% accuracy was confirmed.