FashionCLIP adapts the CLIP vision-language architecture to the fashion domain by fine-tuning on a fashion-specific dataset. It enables zero-shot image classification and feature extraction for fashion products, allowing similarity search and categorization in e-commerce settings. The model uses contrastive learning to align image and text representations in a shared embedding space.