From Snap to Score: A Comprehensive Resource for Predicting Fashion Preferences Using Competitive-Based Ranking

The Niau\footnote{The ``Niau’’ (似合う) is where an outfit enhances the wearer’s visual appeal from a third-party perspective.} Dataset introduces a novel, high-quality collection of 6,000 fashion images curated from the WEAR\footnote{https://wear.jp} Japanese Social Networking Service for Fashion and an in-house studio to analyze fashion preferences among young females in Japan. Utilizing the OpenSkill\cite{Joshy2024} algorithm for real-time online ranking and incorporating innovative data preprocessing techniques, the dataset ensures the integrity of user evaluations. This paper demonstrates the potential of the Niau Dataset in training machine learning models to predict fashion preferences, with the CNN model achieving significant correlation metrics on both test and disjoint datasets. The insights gained from this study have practical applications in the fashion industry, providing valuable data for trend analysis, personalized recommendations, and customer engagement strategies. While the dataset itself is not open-sourced, the methodologies and techniques developed are shared to advance the field of fashion analysis.

The 27th Meeting on Image Recognition and Understanding