Skip to content
#

compositional

Here are 14 public repositories matching this topic...

CLIPIndia 🇮🇳👗 adapts SigLIP for Indian fashion using parameter-efficient contrastive fine-tuning. It enables text, image, and compositional image+text search. Results: +57.7% Recall@1 on regional apparel, <2% degradation on general retrieval, trained efficiently on a single GPU.

  • Updated Jul 25, 2026
  • Jupyter Notebook

Add this topic to your repo

To associate your repository with the compositional topic, visit your repo's landing page and select "manage topics."

Learn more