Nike, Adidas, or Asics? Evolution of sneakers design in 1999-2022 as seen by a neural net
Authors:
Sungkyu Park, Hyeonho Song, Sungwon Han, Berhane Weldegebriel, Lev Manovich, Emanuele Arielli, Meeyoung Cha.
Publication:
Processings of World Wide Web International Conference 2022, April 25–29, 2022, Lyon, France.
Abstract:
Our study for the first time looks at product design evolution over a long period using machine learning methods. It examines 23,492 sneaker images and metadata obtained from StockX.com, a global reselling site. These sneaker designs span 22 years, from 1999 to 2020. We propose a new tool for understanding historical design changes that we call "design index." As used in this paper, design index combines many sneakers' characteristics into a single number. Such indexes can be computed for any brand or types of sneakers, thus allowing us to compare changes in their models design over time. This method can be also used to study historical changes in other cultural fields such as literature, cinema, visual art, architecture, social media photography, etc.
Our study shows how well-known brands have maintained their distinct identities in the "design space" of possible color color combinations and shapes. The model also predicts which sneakers will likely command a high premium in the reselling market, indicating potential investment and design strategies to use in the future.