PIKAR: A Pixel-Level Image Kansei Analysis and Recognition System Based on Deep Learning for User-Centered Product Design

摘要: Can machines learn to perceive products like humans? Kansei Engineering has been developed connect product design and human perception. While conventional Systems exhibit a high dependency on manual extraction of elements thereby are restricted validity issues, we present Pixel-level Image Analysis Recognition (PIKAR) system that applies deep learning extract analyze the formation perception towards designs automatically. Method validation is performed based evaluation cosmetic packaging’s kawaii. Two neural nets trained 1,414 images, labeled by eight participants their kawaii (1–5 Likert Scale), have achieved better prediction than test persons. The implemented neuron analysis methodology for points consistency with previous experimental studies gives insight individual differences. This work addresses possibility applying support user experience researches.