← back to Handbag Authentication

HANDBAG_DATA_SOURCES.md

373 lines

# Comprehensive Handbag Data Sources from GitHub

*Compiled: 2025-11-11*

This document aggregates all available handbag datasets found on GitHub, including historical data from museums, universities, and libraries.

---

## 🎯 Primary Handbag Datasets

### 1. **edges2handbags Dataset** (iGAN Project)
- **Source**: Amazon product images
- **Size**: 137,000 images
- **Format**: Image files (64x64 and 128x128 resolutions)
- **Features**: Edge-detected images + original photos
- **GitHub**: https://github.com/junyanz/iGAN
- **Usage**: GAN training, edge-to-photo translation
- **Data Type**: Computer vision training dataset
- **Related Projects**:
  - https://github.com/carpedm20/DiscoGAN-pytorch
  - https://github.com/junyanz/BicycleGAN
  - https://github.com/kmbae/ImaGAN

### 2. **Pursearch - Branded Handbag Dataset**
- **Creator**: jenniening (Insight Fellow Program)
- **Size**: 5,000+ images
- **Brands**: 6 most popular handbag brands
- **Styles**: 18 different handbag styles
- **GitHub**: https://github.com/jenniening/Pursearch
- **Features**:
  - Branded handbag dataset
  - Retailer handbag dataset (Saks Fifth Avenue, Shein)
  - 101 brands represented
  - 3,000+ styles
- **Data Source**: Google search results scraping
- **Format**: Images with brand/style metadata
- **Use Case**: Image-based handbag identification and classification

### 3. **Luxury Handbag Price Prediction Dataset**
- **Creator**: mcabinaya
- **GitHub**: https://github.com/mcabinaya/Luxury-Handbag-Price-Prediction
- **Brands**: Celine, Gucci, Tom Ford, Louis Vuitton
- **Data Collection Method**: Expert interviews and brand data
- **Features Analyzed**:
  - Brand
  - Skin/material type
  - Number of components
  - Various bag characteristics
  - Pricing data
- **Use Case**: Price prediction, regression analysis
- **Format**: Structured data (CSV/spreadsheet format likely)

### 4. **DeepFashion2 Dataset** (Includes Bags)
- **GitHub**: https://github.com/switchablenorms/DeepFashion2
- **Total Size**: 491,000 diverse images
- **Categories**: 13 popular clothing categories (includes bags)
- **Sources**: Commercial shopping stores + consumer images
- **Format**: JSON annotation files for each image
- **Features**:
  - Bounding boxes
  - Segmentation masks
  - Keypoints
  - Category labels
- **Use Case**: Fashion detection, segmentation, retrieval

---

## 🏛️ Museum & Historical Collections

### 5. **Metropolitan Museum of Art Open Access**
- **GitHub**: https://github.com/metmuseum/openaccess
- **Total Objects**: 492,000+ artworks
- **Costume Institute Collection**: 33,000+ costumes and accessories
  - Includes handbags and purses
  - 5 continents represented
  - 7 centuries of fashion history
- **Format**: CSV (MetObjects.csv via Git LFS)
- **Encoding**: UTF-8
- **Direct Access**: https://github.com/metmuseum/openaccess/blob/master/MetObjects.csv
- **License**: Public domain, unrestricted use
- **API**: Available at https://metmuseum.github.io/
- **Search Fields**:
  - Object Name
  - Department (Costume Institute)
  - Classification
  - Period
  - Culture
  - Medium
  - Dimensions

### 6. **Harvard Art Museums API & Dataset**
- **API Docs**: https://github.com/harvardartmuseums/api-docs
- **Website**: https://harvardartmuseums.org/collections/api
- **Format**: JSON via REST API
- **Harvard Dataverse**: Multiple dataset versions available
  - DOI: 10.7910/DVN/AGTG4E
  - DOI: 10.7910/DVN/BQAQ7G
- **Features**:
  - Aggregation parameters for statistics
  - Filter by object type
  - Includes fashion and accessories
- **API Access**: Requires API key
- **Use Case**: Art historical research, museum data analysis

### 7. **Victoria & Albert Museum (V&A) Collections**
- **API Portal**: https://developers.vam.ac.uk/
- **API Guide**: https://developers.vam.ac.uk/guide/v2/
- **Collection Size**: 1.2+ million objects
- **Images**: 500,000+ images
- **Fashion Search**: https://collections.vam.ac.uk/search/?id_category=THES48957
- **Format**: CSV and JSON export options
- **Features**:
  - Metadata for all objects
  - High-resolution images
  - Fashion and textile collections
  - Handbags and accessories included
- **License**: Varies by object (check V&A terms)

### 8. **National Gallery of Art Open Data**
- **GitHub**: https://github.com/NationalGalleryOfArt/opendata
- **Collection Size**: 130,000+ artworks
- **Format**: CSV/JSON
- **Includes**: Artist information, artwork metadata
- **Note**: May have limited fashion/handbag content

---

## 📊 Fashion Market & Price Analysis Datasets

### 9. **Farfetch Luxury Fashion Dataset**
- **GitHub**: https://github.com/haridev98/A-Data-Driven-Look-at-the-Luxury-Fashion-Market
- **Size**: 100,000+ products
- **Features**:
  - Price
  - Brand
  - Category
  - Country of origin
- **Use Case**: Market analysis, pricing trends

### 10. **E-Commerce Dataset Samples** (Luminati)
- **GitHub**: https://github.com/luminati-io/eCommerce-dataset-samples
- **Content**: Multiple e-commerce dataset samples
- **Size**: 1,000+ records per sample
- **Use Cases**:
  - Product trend analysis
  - Pricing strategies
  - Consumer sentiment insights

### 11. **Luxury Shop Revenue Analytics**
- **GitHub**: https://github.com/tesserakh/luxury-shop
- **Features**:
  - Product name
  - Price
  - Discount
  - Availability
- **Use Case**: Revenue analytics for luxury bags and accessories

---

## 🎓 University & Academic Datasets

### 12. **University of Illinois Datasets**
- **GitHub**: https://github.com/wadefagen/datasets
- **Content**: Various interesting datasets from UIUC
- **Note**: General academic datasets repository

### 13. **HistData - Historical Statistics**
- **GitHub**: https://github.com/friendly/HistData
- **Content**: Data Sets from the History of Statistics and Data Visualization
- **Format**: R data packages
- **Note**: May contain historical fashion/consumer data

---

## 🔍 Additional Resources

### Fashion Dataset Topics on GitHub
- **Topic Page**: https://github.com/topics/fashion-dataset
- **Handbag Topic**: https://github.com/topics/handbag
- **Museum Collections**: https://github.com/topics/museum-collections

### Related Clothing Datasets

#### Clothing Detection Dataset
- **GitHub**: https://github.com/alexeygrigorev/clothing-dataset
- **Size**: Multiple classes of clothing items
- **Format**: Images with annotations

#### Fashion-MNIST
- **GitHub**: https://github.com/ymattu/fashion-mnist-csv
- **Format**: CSV
- **Size**: Standard MNIST size (60K training, 10K test)
- **Note**: Grayscale thumbnails, limited for handbag specifics

---

## 📥 Download Instructions

### For Git LFS Files (e.g., Met Museum CSV)
```bash
# Clone with LFS support
git lfs install
git clone https://github.com/metmuseum/openaccess.git
cd openaccess
git lfs pull
```

### For Large Datasets
Most large image datasets require:
1. Filling out request forms
2. Agreeing to terms of use
3. Academic/research verification
4. Direct download links provided via email

### API Access (Harvard, V&A)
```bash
# Example API call structure
curl "https://api.harvardartmuseums.org/object?apikey=YOUR_KEY&classification=Handbags"
curl "https://api.vam.ac.uk/v2/objects/search?q=handbag"
```

---

## 🎯 Recommended Datasets by Use Case

### For Machine Learning / Computer Vision
1. **edges2handbags** (137K images) - Best for GANs
2. **DeepFashion2** (491K images) - Best for detection/segmentation
3. **Pursearch** (5K+ images) - Best for brand/style classification

### For Price Analysis / Market Research
1. **Luxury Handbag Price Prediction** - Expert-curated pricing data
2. **Farfetch Dataset** - 100K+ luxury products with pricing
3. **Luxury Shop Revenue Analytics** - E-commerce transaction data

### For Historical Research
1. **Met Museum Open Access** - 7 centuries of fashion history
2. **V&A Collections** - Extensive historical fashion archive
3. **Harvard Art Museums** - Academic research-focused

### For E-Commerce Applications
1. **Pursearch** - Retailer data with brands and styles
2. **E-Commerce Dataset Samples** - Multi-source product data
3. **edges2handbags** - Large-scale product images

---

## 📊 Data Statistics Summary

| Dataset | Records | Format | Historical | Price Data | Images | API |
|---------|---------|--------|------------|------------|--------|-----|
| edges2handbags | 137K | Images | No | No | ✅ | No |
| Pursearch | 5K+ | Images | No | Limited | ✅ | No |
| Luxury Price Pred | <1K | Structured | No | ✅ | Limited | No |
| DeepFashion2 | 491K | JSON+Images | No | No | ✅ | No |
| Met Museum | 492K+ | CSV | ✅ | No | ✅ | ✅ |
| Harvard Museums | Variable | JSON | ✅ | No | ✅ | ✅ |
| V&A | 1.2M+ | CSV/JSON | ✅ | No | ✅ | ✅ |
| Farfetch | 100K+ | Structured | No | ✅ | Limited | No |

---

## 🔗 Quick Links

### GitHub Repositories
- **Met Museum**: https://github.com/metmuseum/openaccess
- **Harvard Museums**: https://github.com/harvardartmuseums/api-docs
- **iGAN/edges2handbags**: https://github.com/junyanz/iGAN
- **Pursearch**: https://github.com/jenniening/Pursearch
- **DeepFashion2**: https://github.com/switchablenorms/DeepFashion2
- **Luxury Price**: https://github.com/mcabinaya/Luxury-Handbag-Price-Prediction

### API Endpoints
- **Met Museum API**: https://metmuseum.github.io/
- **Harvard API**: https://harvardartmuseums.org/collections/api
- **V&A API**: https://developers.vam.ac.uk/

### Data Portals
- **Harvard Dataverse**: https://dataverse.harvard.edu/
- **Met Open Access**: https://www.metmuseum.org/about-the-met/policies-and-documents/open-access
- **V&A Collections**: https://collections.vam.ac.uk/

---

## 📝 Notes on Data Quality

### Strengths by Source

**Museum Collections (Met, Harvard, V&A)**
- ✅ High-quality metadata
- ✅ Historical provenance
- ✅ Expert curation
- ✅ Multiple centuries of data
- ⚠️ May have gaps in modern/contemporary items

**Computer Vision Datasets (edges2handbags, DeepFashion2)**
- ✅ Large scale (100K+ images)
- ✅ Consistent formatting
- ✅ Pre-processed annotations
- ⚠️ Limited metadata beyond visual features

**E-Commerce/Price Datasets**
- ✅ Current market data
- ✅ Pricing information
- ✅ Brand information
- ⚠️ May lack historical depth
- ⚠️ Limited to modern products

---

## 🚀 Getting Started

### Step 1: Identify Your Use Case
- Academic research → Museum APIs
- ML/CV development → edges2handbags or DeepFashion2
- Price analysis → Luxury Price Prediction or Farfetch
- Brand classification → Pursearch

### Step 2: Access Requirements
- Public datasets: Direct download
- Museum APIs: Register for API key
- Academic datasets: May require institutional affiliation

### Step 3: Data Processing
- CSV files: Use pandas, Excel, or database import
- Images: Requires storage (10GB+ for large datasets)
- APIs: Rate limits apply, plan accordingly

---

## 📧 Data Access Contacts

Most datasets are freely available, but some require:
- **Harvard Art Museums**: API key via https://harvardartmuseums.org/collections/api
- **V&A**: API access via https://developers.vam.ac.uk/
- **DeepFashion2**: May require academic verification
- **Commercial datasets**: Check individual repository licenses

---

## ⚖️ License Information

**Public Domain / Open Access:**
- Met Museum Open Access
- National Gallery of Art

**API Terms Required:**
- Harvard Art Museums
- V&A Collections

**Academic/Research Use:**
- DeepFashion2
- Some university datasets

**Check Individual Licenses:**
- GitHub repositories (usually MIT or Apache 2.0)
- E-commerce datasets (may have restrictions)

---

## 🔄 Last Updated
2025-11-11

## 📌 Maintenance
This document should be updated periodically as new datasets become available. Check GitHub topics for latest additions:
- https://github.com/topics/handbag
- https://github.com/topics/fashion-dataset
- https://github.com/topics/museum-collections

---

*End of Document*