GCSP research stipend | Fall 2025

Fine-tuning Garment Digitization Models for Texture Recognition

Data icon, disabled. Four grey bars arranged like a vertical bar chart.

AI-powered garment digitization enables rapid conversion of real-world clothing into digital counterparts. Current models capture basic garment structures but cannot infer surface-level details and material textures, critical properties for higher-fidelity reconstruction. This research aims to improve this through targeted fine-tuning, extending the JSON configuration of existing programming frameworks used to design parametric sewing patterns. To train the model, the researcher will use a 50-garment dataset with unique textures and heavy surface detail. The researcher hypothesizes that this will improve texture and surface detail recognition while maintaining structural quality, thus enabling a greater focus on creative work and rapid design iteration.