🛍️ GenAI Fashion Store

An AI-powered fashion shopping assistant that uses computer vision, natural language processing, and recommendation algorithms to provide a personalized shopping experience. Built with Streamlit, OpenAI's CLIP, and GPT models.

✨ Features

🔍 Smart Search Capabilities

• Text-based Search: Find products using natural language queries
• Visual Search: Upload an image to find similar fashion items
• Hybrid Search: Combine text and image search for precise results
• Advanced Filtering: Filter by gender, category, color, season, and usage

🤖 AI Shopping Assistant

• Natural language interaction powered by GPT
• Style advice and fashion tips
• Outfit coordination suggestions
• Product information and recommendations

🎯 Intelligent Recommendations

• Visual Similarity: Find products with similar visual features using CLIP embeddings
• Category-based: Discover items from related categories
• Collaborative Filtering: Get recommendations based on user behavior
• Hybrid Approach: Combines multiple strategies for best results
• Complementary Items: Find accessories and items that complete an outfit

📊 Advanced Backend

• Vector database (ChromaDB) for efficient similarity search
• CLIP embeddings for visual feature extraction
• Caching system for improved performance
• Batch processing for large datasets

🚀 Quick Start

Prerequisites

• Python 3.8 - 3.12 (Note: Python 3.13 may have compatibility issues)
• 4GB+ free disk space
• 8GB+ RAM recommended
• CUDA-capable GPU (optional, for better performance)

Installation

1. Clone the repository

git clone https://github.com/codeWithUtkarsh/GenAIFashionStore
cd GenAIFashionStore

2. Create and activate virtual environment

# Create virtual environment
python3 -m venv .venv

# Activate virtual environment
source .venv/bin/activate  # On macOS/Linux
# OR
.venv\Scripts\activate  # On Windows

3. Install dependencies

For stable installation (recommended):

# Use stable requirements to avoid compatibility issues
pip install -r requirements_stable.txt

For latest versions:

pip install -r requirements.txt

4. Run the setup script

python setup.py

This will:

• Verify your Python version
• Set up configuration files
• Create necessary directories
• Guide you through API key configuration

5. Configure API Keys

Edit the .env file with your API keys:

OPENAI_API_KEY=your_openai_api_key_here
KAGGLE_USERNAME=your_kaggle_username
KAGGLE_KEY=your_kaggle_api_key

Get your keys from:

• OpenAI: https://platform.openai.com/api-keys
• Kaggle: https://www.kaggle.com/settings/account

Running the Application

Safe Start (Recommended for First Run)

# Use the safe runner script to handle common issues
python run_safe.py --clear-cache

Normal Start

streamlit run app.py

Custom Options

# Run on different port
python run_safe.py --port 8502

# Enable debug mode
python run_safe.py --debug

# Clear all caches before starting
python run_safe.py --clear-cache

5. Open in browser Navigate to: http://localhost:8501

📁 Project Structure

GenAIFashionStore/
│
├── app.py                  # Main Streamlit application
├── config.py               # Configuration settings
├── setup.py                # Setup script
├── requirements.txt        # Python dependencies
├── .env.example           # Environment variables template
│
├── data_downloader.py      # Kaggle dataset downloader
├── image_embedder.py       # CLIP-based image embedding
├── vector_database.py      # ChromaDB vector storage
├── recommendation_engine.py # Multi-strategy recommendations
├── genai_assistant.py      # GPT-powered chat assistant
│
├── data/                   # Dataset directory
│   ├── images/            # Product images
│   └── styles.csv         # Product metadata
│
├── models/                 # Saved models
├── vector_db/             # Vector database storage
├── chroma_db/             # ChromaDB persistence
└── .cache/                # Embedding cache

🎯 Usage Guide

Searching for Products

1. Text Search: Type queries like:

   • "blue shirt for men"
   • "summer dresses"
   • "formal shoes"

2. Visual Search: Upload an image of a fashion item to find similar products

3. Filters: Use sidebar filters to narrow down results by:

   • Gender
   • Category
   • Color
   • Season
   • Usage/Occasion

Using the AI Assistant

Ask questions like:

• "What goes well with blue jeans?"
• "Show me outfit ideas for a summer wedding"
• "I need business casual recommendations"
• "What's trending this season?"

Getting Recommendations

1. Click on any product to view details
2. System automatically shows:
   • Similar items
   • Complementary pieces
   • Complete outfit suggestions

🔧 Configuration

Adjusting Dataset Size

Edit config.py to change the number of products loaded:

MAX_PRODUCTS_TO_LOAD = 2000  # Adjust this value

Model Selection

Configure embedding models in config.py:

EMBEDDING_MODEL = "clip-ViT-B-32"  # Options: clip-ViT-B-32, clip-ViT-L-14
GPT_MODEL = "gpt-4-turbo-preview"  # Options: gpt-3.5-turbo, gpt-4

Performance Tuning

# Enable GPU acceleration (if available)
USE_GPU = True

# Adjust batch size for processing
BATCH_SIZE = 32

# Configure cache settings
ENABLE_CACHE = True
CACHE_TTL = 3600  # Cache duration in seconds

🗃️ Dataset

This project uses the Fashion Product Images (Small) dataset from Kaggle:

• Source: https://www.kaggle.com/datasets/paramaggarwal/fashion-product-images-small
• Size: ~44,000 products
• Contents: Product images, categories, attributes, and metadata

Dataset Features

• Product images in multiple categories
• Detailed metadata (gender, category, color, season, usage)
• Structured CSV format for easy processing

🛠️ Technical Stack

Core Technologies

• Frontend: Streamlit
• Backend: Python 3.8-3.12
• Database: ChromaDB (Vector Database)
• AI/ML: OpenAI CLIP, GPT, Sentence Transformers

Key Libraries

• streamlit: Web application framework
• torch & torchvision: Deep learning
• transformers: Hugging Face models
• chromadb: Vector database
• openai: GPT integration
• Pillow: Image processing
• pandas & numpy: Data manipulation

📊 Performance Considerations

System Requirements

• Minimum: 4GB RAM, 2 CPU cores
• Recommended: 8GB RAM, 4+ CPU cores, CUDA GPU
• Storage: 5-10GB for dataset and models

🚀 Advanced Features

Adding Custom Models

1. Implement embedding interface in image_embedder.py
2. Update config.py with model configuration
3. Modify vector database dimensions if needed

Extending Recommendations

Add new recommendation strategies in recommendation_engine.py:

def _custom_recommendations(self, ...):
    # Your custom logic here
    pass

📈 Future Enhancements

• User authentication and profiles
• Purchase history tracking
• Size and fit recommendations
• Virtual try-on using AR
• Multi-language support
• Price tracking and alerts
• Social features (sharing, reviews)
• Mobile app version
• Integration with e-commerce platforms

Made with ❤️ for AffinityLabs