Fashion Recommendation Engine

A high-performance, content-based fashion recommendation system that suggests similar products using multimodal AI (text + image). Built with FastAPI, CLIP, SBERT, and FAISS.

🌐 Live Demo: https://fashion-recommendation-engine.onrender.com/app

Note: The demo is hosted on Render's free tier. It may take 30-60 seconds to wake up after 15 minutes of inactivity. Please be patient! ⏳

🎥 Demo

Direct link: fasrec_project_demo.mp4

✨ Features

• Multimodal Search: Recommendations based on both visual similarity (CLIP) and semantic meaning (SBERT).
• Blazing Fast: Uses FAISS HNSW index for millisecond-latency nearest neighbor search on 44k+ products.
• Smart Recommendations: Hybrid fusion of text and image scores (α * text_sim + (1-α) * image_sim) for highly relevant results.
• Modern UI: Dark-themed, responsive frontend with glassmorphism design.
• Cloud-Native: Images served via Cloudflare R2 CDN, app deployed on Render.

🛢 Dataset

• Link: https://www.kaggle.com/datasets/paramaggarwal/fashion-product-images-small

🛠️ Tech Stack

• Backend: Python 3.11, FastAPI, Uvicorn
• AI/ML: OpenAI CLIP (Vision), SBERT (Metadata), FAISS (Vector Search)
• Frontend: HTML5, CSS3, Vanilla JS
• Infrastructure: Docker, Render, Cloudflare R2

📦 Project Structure

├── artifacts/          # Precomputed embeddings & FAISS indexes
├── data/              # Dataset (styles.csv)
├── frontend/          # UI logic (index.html)
├── scripts/           # ML pipelines (embedding gen, indexing)
└── src/               # Application source code

🔧 Local Development

Prerequisites

• Python 3.11+
• Docker (optional)

Setup

1. Clone the repository:

   git clone https://github.com/HimasagarU/Fashion-Recommendation-Engine.git
   cd Fashion-Recommendation-Engine

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   python -m uvicorn src.api:app --reload

   Open http://localhost:8000/app in your browser.

Using Docker

docker-compose up --build

📝 License

This project is open source and available under the MIT License.