Met gala classifier
Introduction
The Met Gala Classifier is a machine learning project that attempts to automate something subjective: deciding whether a red carpet outfit actually fits a specific museum exhibition theme. We focused on the Metropolitan Museum of Art’s “Heavenly Bodies: Fashion and the Catholic Imagination” exhibition, curating a dataset of historical and contemporary looks to train our models.
Setup
From a technical standpoint, the pipeline relies on a combination of a Convolutional Neural Network and a Support Vector Machine model. The CNN handles feature extraction from high-resolution images of the garments, picking up on textures, silhouettes, and stylistic patterns that reference religious iconography.Those extracted feature vectors are then passed to the SVM model, which performs the final classification to determine whether an outfit belongs in the “on-theme” or “off-theme” category. Curating clean exhibition data was crucial to keeping the model from latching onto background noise or lighting differences rather than the actual design elements.