AI Tailoring: Evaluating Influence of Image Features on Fashion Product Popularity

TL;DR

A robust methodology to ascertain the most impactful features in fashion product images, utilizing past market sales data is introduced and a metric called "influence score" is proposed to quantitatively assess the importance of product features.

Abstract

Identifying key product features that influence consumer preferences is essential in the fashion industry. In this study, we introduce a robust methodology to ascertain the most impactful features in fashion product images, utilizing past market sales data. First, we propose the metric called "influence score" to quantitatively assess the importance of product features. Then we develop a forecasting model, the Fashion Demand Predictor (FDP), which integrates Transformer-based models and Random Forest to predict market popularity based on product images. We employ image-editing diffusion models to modify these images and perform an ablation study, which validates the impact of the highest and lowest-scoring features on the model's popularity predictions. Additionally, we further validate these results through surveys that gather human rankings of preferences, confirming the accuracy of the FDP model's predictions and the efficacy of our method in identifying influential features. Notably, products enhanced with "good" features show marked improvements in predicted popularity over their modified counterparts. Our approach develops a fully automated and systematic framework for fashion image analysis that provides valuable guidance for downstream tasks such as fashion product design and marketing strategy development.

Figures (7)

• Figure 1: Pipeline of our methodology
• Figure 2: Density distribution of normalized fashion product sales. The sales are normalized between 0 and 1 using min-max scaling, showing a skewed distribution with a higher concentration of lower sales values. The red dashed lines indicate thresholds that divide the data into three equal quantiles.
• Figure 3: A schematic overview o Fashion Demand Predictor(FDP): the model takes three types of input features: (1) descriptive text features processed by the Sentence-BERT's encoder, (2) image features processed by the FashionCLIP's encoder, and (3) one-hot encoded categorical and numeric features. The encoded descriptive and image features are further compressed by an autoencoder.
• Figure 4: Comparison of Original (left) and AI-Modified (right) Designs. The AI-Modified design demonstrates the effect of removing a "Good" feature Folded Cuffs.
• Figure 5: Comparison of Original (left) and AI-Modified (right) Designs. The AI-Modified design demonstrates the effect of removing a "Good" feature Adjustable Inner Drawstring Waist.