Multi-Modal Semantic Image Search Engine
Open-source semantic image search supporting text→image, image→image, and text+image→image queries on the Myntra Fashion Product Dataset.
Problem
Fashion catalogs need flexible search beyond text or image alone — shoppers search by photo, description, or both.
Approach
Joint embedding space via multimodal encoders and FAISS index for sub-second retrieval across query modalities.
Architecture
Catalog → image + text encoders → unified vector index → multi-modal query API → Streamlit demo UI.
Results
Real-time text-to-image, image-to-image, and text+image retrieval on the Myntra Fashion dataset.
Lessons learned
Late-fusion of modalities at query time is simpler and competitive with joint training for many retrieval tasks.
Next
Fed-BLEND — Federated Conformal Prediction for VLMs
Novel federated conformal prediction method that mitigates hallucinations in federated fine-tuned vision-language models via abstention.
GCTAF — Time-Series Forecasting & Flare Risk Classification
Attention-based forecasting of magnetic field trajectories combined with supervised contrastive learning for solar flare classification.