From 2ac8f6acb0f21733f6d937862116035b2996abeb Mon Sep 17 00:00:00 2001 From: Arya Sahu <121129315+kirbynuggets@users.noreply.github.com> Date: Mon, 21 Apr 2025 13:07:07 +0530 Subject: [PATCH 1/4] Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 084b3671..0e30ddf7 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,3 @@ -# cs331 +# Outfit-Recommendation System -Outfit recommendation system as part of the lab project for CS331, Winter 2024-2025. +An E-commerce website with a recommender system as part of the lab project for CS331, Winter 2024-2025. From 1ed8ad45b92b598170c32a066d99a8d78e35680d Mon Sep 17 00:00:00 2001 From: Arya Sahu <121129315+kirbynuggets@users.noreply.github.com> Date: Sat, 31 May 2025 18:47:10 +0530 Subject: [PATCH 2/4] Update README.md --- README.md | 108 +++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 106 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 0e30ddf7..93f485ec 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,107 @@ -# Outfit-Recommendation System +# RAG: Outfit Recommendation System -An E-commerce website with a recommender system as part of the lab project for CS331, Winter 2024-2025. +**AI-Driven Outfit Recommendations with Advanced Search Capabilities** + +*Department of Computer Science and Engineering, Indian Institute of Information Technology Guwahati* + +## Abstract + +This project delivers a full-stack fashion e-commerce platform that implements advanced outfit recommendations and sophisticated search. Our solution employs architecture with a **Retrieval-Augmented Generation (RAG)** framework that employs state-of-the-art machine learning, including **OpenAI CLIP** for multimodal understanding and **Maximal Marginal Relevance (MMR)** for diverse recommendations. + +## 🎯 Motivation + +Empower users to discover and explore outfit ideas tailored to their searches, using AI-driven image and text understanding, moving beyond sales to genuine style inspiration and intuitive interaction. + +## 🏗️ Architectural Approach + +### Microservices Architecture +- **Scalable Design**: Breaking the platform into independent services +- **Agility**: Platform design supports parallel development and rapid feature deployment +- **Quick Iterations**: Allowing quick tweaks to features and search algorithms + +### Scalability & Resilience +- **Loose Coupling**: Ensures easy integration of components +- **High Availability**: Robust system design for consistent performance + +## 🚀 Key Features + +### Advanced Search Capabilities +- **RAG Architecture Integration**: Enhances search with information retrieval from the product database +- **Multimodal Embeddings (OpenAI CLIP)**: Generate unified image and text embeddings for deep semantic understanding +- **Cross-Modal Search**: Enable attribute prediction (e.g., garment, article type, color) and cross-modal search functionality + +### Intelligent Recommendation Engine +- **TF-IDF**: Weighs textual feature importance for content-based recommendations +- **One-Hot Encoding**: Encodes categorical features for similarity calculations +- **Similarity Search (Annoy)**: Fast Approximate Nearest Neighbors (ANN) search using cosine similarity for real-time recommendations + +### Recommendation Strategy +- **MMR (Maximal Marginal Relevance)**: Balances relevance and diversity in recommendations +- **K-Means Clustering**: Detects dominant colors for outfit compatibility scoring + +## 📊 Performance Results + +Our system demonstrates significant improvements over baseline models: + +| Metric | Value | Improvement | +|--------|-------|-------------| +| NDCG@5 | 0.991 | 15.37% | +| Novelty | 0.947 | 2.85% | + +### Industry Comparison +The system shows superior performance compared to existing fashion e-commerce platforms in key areas: +- **Multimodal Search**: Enhanced accuracy in product discovery +- **Personalization**: Improved user experience through AI-driven recommendations +- **Search Quality**: Better semantic understanding and relevance + +## 🛠️ Technical Stack + +### Machine Learning & AI +- **OpenAI CLIP**: Multimodal embeddings for image-text understanding +- **TF-IDF**: Text feature importance weighting +- **Annoy**: Approximate Nearest Neighbors for fast similarity search +- **K-Means**: Color clustering for compatibility analysis + +### Architecture Components +- **RAG Framework**: Retrieval-Augmented Generation for enhanced search +- **Microservices**: Scalable and maintainable service architecture +- **MMR Algorithm**: Maximal Marginal Relevance for diverse recommendations + +## 🎯 Core Capabilities + +1. **Smart Product Discovery**: AI-driven search that understands both images and text queries +2. **Personalized Recommendations**: Tailored outfit suggestions based on user preferences +3. **Cross-Modal Understanding**: Search using images to find similar or complementary items +4. **Color Compatibility**: Intelligent color matching for outfit coordination +5. **Real-Time Performance**: Fast recommendation engine for seamless user experience + +## 📈 System Benefits + +- **Enhanced User Experience**: Smarter, more personalized outfit discovery +- **Improved Search Accuracy**: 15.37% improvement in NDCG@5 scores +- **Scalable Architecture**: Microservices design supports growth and feature expansion +- **AI-Powered Insights**: Deep understanding of fashion preferences and trends + +## 🔍 Use Cases + +- **Style Discovery**: Help users find new fashion styles and outfit combinations +- **Visual Search**: Search for products using images instead of text +- **Outfit Coordination**: Suggest complementary items for complete outfits +- **Personalized Shopping**: Tailored recommendations based on individual preferences +- **Color Matching**: Find items that work well together color-wise + +## 📚 References + +1. Radford, A., et al. "Learning Transferable VIM From NL Supervision." ICML, 2021. +2. Bernhardsson, E. "Annoy: Approximate Nearest Neighbors in C++/Python." GitHub Repo, 2023. + +## 👥 Contributors + +**Ahlad Pataparla, Anushka Srivastava, Arya Sahu, Khushi Mandal** + +Department of Computer Science and Engineering +Indian Institute of Information Technology Guwahati + +--- + +*This project represents cutting-edge research in AI-driven fashion e-commerce, combining advanced machine learning techniques with practical applications for enhanced user experience in online fashion retail.* From 6cedc4e618b0be410c79b78ac2bab61ea0b7d8c5 Mon Sep 17 00:00:00 2001 From: Arya Sahu <121129315+kirbynuggets@users.noreply.github.com> Date: Sat, 31 May 2025 18:55:02 +0530 Subject: [PATCH 3/4] Update README.md --- README.md | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/README.md b/README.md index 538c8fb8..eb943fc1 100644 --- a/README.md +++ b/README.md @@ -2,8 +2,6 @@ **AI-Driven Outfit Recommendations with Advanced Search Capabilities** -*Department of Computer Science and Engineering, Indian Institute of Information Technology Guwahati* - ## Abstract This project delivers a full-stack fashion e-commerce platform that implements advanced outfit recommendations and sophisticated search. Our solution employs architecture with a **Retrieval-Augmented Generation (RAG)** framework that employs state-of-the-art machine learning, including **OpenAI CLIP** for multimodal understanding and **Maximal Marginal Relevance (MMR)** for diverse recommendations. @@ -273,4 +271,4 @@ LUXE is built with a mobile-first approach, ensuring a seamless experience acros --- -© 2025 LUXE Fashion Marketplace. All rights reserved. \ No newline at end of file +© 2025 LUXE Fashion Marketplace. All rights reserved. From cd4c939084495ac7f814f0e8680b53add59484f4 Mon Sep 17 00:00:00 2001 From: Arya Sahu <121129315+kirbynuggets@users.noreply.github.com> Date: Sun, 1 Jun 2025 22:13:36 +0530 Subject: [PATCH 4/4] Update README.md --- README.md | 105 ------------------------------------------------------ 1 file changed, 105 deletions(-) diff --git a/README.md b/README.md index eb943fc1..b6156502 100644 --- a/README.md +++ b/README.md @@ -1,108 +1,3 @@ -# RAG: Outfit Recommendation System - -**AI-Driven Outfit Recommendations with Advanced Search Capabilities** - -## Abstract - -This project delivers a full-stack fashion e-commerce platform that implements advanced outfit recommendations and sophisticated search. Our solution employs architecture with a **Retrieval-Augmented Generation (RAG)** framework that employs state-of-the-art machine learning, including **OpenAI CLIP** for multimodal understanding and **Maximal Marginal Relevance (MMR)** for diverse recommendations. - -## 🎯 Motivation - -Empower users to discover and explore outfit ideas tailored to their searches, using AI-driven image and text understanding, moving beyond sales to genuine style inspiration and intuitive interaction. - -## 🏗️ Architectural Approach - -### Microservices Architecture -- **Scalable Design**: Breaking the platform into independent services -- **Agility**: Platform design supports parallel development and rapid feature deployment -- **Quick Iterations**: Allowing quick tweaks to features and search algorithms - -### Scalability & Resilience -- **Loose Coupling**: Ensures easy integration of components -- **High Availability**: Robust system design for consistent performance - -## 🚀 Key Features - -### Advanced Search Capabilities -- **RAG Architecture Integration**: Enhances search with information retrieval from the product database -- **Multimodal Embeddings (OpenAI CLIP)**: Generate unified image and text embeddings for deep semantic understanding -- **Cross-Modal Search**: Enable attribute prediction (e.g., garment, article type, color) and cross-modal search functionality - -### Intelligent Recommendation Engine -- **TF-IDF**: Weighs textual feature importance for content-based recommendations -- **One-Hot Encoding**: Encodes categorical features for similarity calculations -- **Similarity Search (Annoy)**: Fast Approximate Nearest Neighbors (ANN) search using cosine similarity for real-time recommendations - -### Recommendation Strategy -- **MMR (Maximal Marginal Relevance)**: Balances relevance and diversity in recommendations -- **K-Means Clustering**: Detects dominant colors for outfit compatibility scoring - -## 📊 Performance Results - -Our system demonstrates significant improvements over baseline models: - -| Metric | Value | Improvement | -|--------|-------|-------------| -| NDCG@5 | 0.991 | 15.37% | -| Novelty | 0.947 | 2.85% | - -### Industry Comparison -The system shows superior performance compared to existing fashion e-commerce platforms in key areas: -- **Multimodal Search**: Enhanced accuracy in product discovery -- **Personalization**: Improved user experience through AI-driven recommendations -- **Search Quality**: Better semantic understanding and relevance - -## 🛠️ Technical Stack - -### Machine Learning & AI -- **OpenAI CLIP**: Multimodal embeddings for image-text understanding -- **TF-IDF**: Text feature importance weighting -- **Annoy**: Approximate Nearest Neighbors for fast similarity search -- **K-Means**: Color clustering for compatibility analysis - -### Architecture Components -- **RAG Framework**: Retrieval-Augmented Generation for enhanced search -- **Microservices**: Scalable and maintainable service architecture -- **MMR Algorithm**: Maximal Marginal Relevance for diverse recommendations - -## 🎯 Core Capabilities - -1. **Smart Product Discovery**: AI-driven search that understands both images and text queries -2. **Personalized Recommendations**: Tailored outfit suggestions based on user preferences -3. **Cross-Modal Understanding**: Search using images to find similar or complementary items -4. **Color Compatibility**: Intelligent color matching for outfit coordination -5. **Real-Time Performance**: Fast recommendation engine for seamless user experience - -## 📈 System Benefits - -- **Enhanced User Experience**: Smarter, more personalized outfit discovery -- **Improved Search Accuracy**: 15.37% improvement in NDCG@5 scores -- **Scalable Architecture**: Microservices design supports growth and feature expansion -- **AI-Powered Insights**: Deep understanding of fashion preferences and trends - -## 🔍 Use Cases - -- **Style Discovery**: Help users find new fashion styles and outfit combinations -- **Visual Search**: Search for products using images instead of text -- **Outfit Coordination**: Suggest complementary items for complete outfits -- **Personalized Shopping**: Tailored recommendations based on individual preferences -- **Color Matching**: Find items that work well together color-wise - -## 📚 References - -1. Radford, A., et al. "Learning Transferable VIM From NL Supervision." ICML, 2021. -2. Bernhardsson, E. "Annoy: Approximate Nearest Neighbors in C++/Python." GitHub Repo, 2023. - -## 👥 Contributors - -**Ahlad Pataparla, Anushka Srivastava, Arya Sahu, Khushi Mandal** - -Department of Computer Science and Engineering -Indian Institute of Information Technology Guwahati - ---- - -*This project represents cutting-edge research in AI-driven fashion e-commerce, combining advanced machine learning techniques with practical applications for enhanced user experience in online fashion retail.* # LUXE - Premium Fashion E-Commerce Platform