The Shape of Design History: Exploring Evolution of Sneakers Design at Scale Using Neural Embedding

EasyChair Preprint 5993

Abstract

We have crawled sneaker images/metadata online and constructed a deep-learning-based sneaker design embedding model by utilizing three attributes, shape, color, and segment. Our work can be linked with cultural analytics to discover the patterns of "latent" features driving cultural evolution over a long period in areas such as fashion.

Keyphrases: Contrastive Learning, Global fashion trends, Neural-net embedding, Sneaker design, deep representation learning, unsupervised clustering

@booklet{EasyChair:5993,
  author    = {Sungkyu Park and Hyeonho Song and Sungwon Han and Lev Manovich and Emanuele Arielli and Meeyoung Cha},
  title     = {The Shape of Design History: Exploring Evolution of Sneakers Design at Scale Using Neural Embedding},
  howpublished = {EasyChair Preprint 5993},
  year      = {EasyChair, 2021}}