Multi-Modal-AI-search

🧠 Fashion Product Image Search Engine

This project is a multi-modal product search engine that allows users to input a text querry or an image query and retrieves the top matching fashion product images from a dataset using deep learning and vector similarity.

It uses CLIP (Contrastive Language–Image Pre-training) for creating embeddings from both images and text, and FAISS for efficient nearest neighbor search.

Real world applications:

1. E-Commerce Product Search
2. Visual Search (Search by Image or Text)
3. Digital Asset Management (DAM)
4. Content Moderation & Compliance
5. Game Asset or 3D Model Search

---

📁 Dataset

We use a subset of the Fashion Product Images (Small) dataset from Kaggle, consisting of:

• ~15,000 product images
• styles.csv: CSV file mapping product metadata
• styles/: Folder containing per-image JSON files with category, subcategory, and descriptions

querry: red shirts for men: Output:top 5 matching images:

querry:

Oytput: