Estimating Impressions for Clothing, Landscape, and Indoor Images Using CNN

In the fashion industry, development and sale of goods is being carried out based on the brand image and impressions such as “girly” or “elegant.” However, the impressions and brand image vary depending on the recipient; therefore, there is a need for a system to quantitatively evaluate the impressions of people. In this study, we use two datasets that include clothing images with impression attributes, and landscape and indoor images according to the brand image. We developed a system to evaluate the impression using convolutional neural network and discuss the results.