AIAnastasia's picture
Update README with full documentation and metadata
80a4599 verified
|
Raw
History Blame Contribute Delete
8.47 kB
metadata
language:
  - en
  - ru
license: cc-by-4.0
size_categories:
  - 1K<n<10K
task_categories:
  - image-classification
  - text-generation
  - object-detection
pretty_name: Georgian Attractions Dataset
tags:
  - georgia
  - tourism
  - attractions
  - travel
  - cultural-heritage
  - geography
  - landmarks
  - multilingual
dataset_info:
  features:
    - name: id
      dtype: int64
    - name: name
      dtype: string
    - name: description
      dtype: string
    - name: category
      dtype: string
    - name: location
      dtype: string
    - name: tags
      dtype: string
    - name: language
      dtype: string
    - name: photo_name
      dtype: string
    - name: license
      dtype: string
    - name: photo_author
      dtype: string
    - name: image
      dtype: image
  splits:
    - name: train
      num_bytes: 2283454828
      num_examples: 1715
  download_size: 1116813006
  dataset_size: 2283454828

Georgian Attractions Dataset 🇬🇪

A comprehensive bilingual dataset featuring 1,715 Georgian tourist attractions with 1,522 high-quality images, descriptions in Russian and English, and detailed metadata including location, category, and licensing information.

Dataset Description

This dataset provides extensive information about tourist attractions, landmarks, and points of interest across Georgia. It includes national parks, museums, fortresses, monasteries, natural landmarks, and historical sites. Each entry contains bilingual descriptions, geographic location data, categorization, searchable tags, and properly licensed images with attribution.

The dataset is designed to support various applications including tourism technology, computer vision research, multilingual NLP tasks, and cultural heritage preservation projects.

Key Features

  • Bilingual Content: Parallel descriptions in Russian and English for most attractions
  • High-Quality Images: 1,522 properly licensed images with full attribution
  • Rich Metadata: Detailed location information, categories, and searchable tags
  • Diverse Coverage: Multiple attraction types from natural parks to historical monuments
  • Clean Structure: Well-organized fields suitable for machine learning applications

Dataset Statistics

Metric Value
Total Records 1,715
Records with Images 1,522 (89%)
Records without Images 193 (11%)
Languages Russian, English
Average Description Length 150-250 words
Image Formats JPG, JPEG, PNG

Dataset Structure

Data Fields

  • id (int64): Unique identifier for each attraction
  • name (string): Name of the attraction in the record's language
  • description (string): Detailed description of the attraction, its history, and significance
  • category (string): Type of attraction (e.g., "National Park", "Museum", "Fortress", "Monastery")
  • location (string): Geographic location within Georgia (region, municipality)
  • tags (string): Comma-separated searchable keywords
  • language (string): Language code of the record ("ru" or "en")
  • photo_name (string): Filename of the associated image
  • license (string): License information for the image (Creative Commons, Public Domain, etc.)
  • photo_author (string): Name of the photographer or image creator
  • image (Image): PIL Image object of the attraction

Data Splits

The dataset contains a single split:

  • train: 1,715 examples

Usage

Loading the Dataset

from datasets import load_dataset

# Load the full dataset
dataset = load_dataset("{repo_id}")

# Access the training split
train_data = dataset['train']

# View a single example
example = train_data[0]
print(f"Name: {{example['name']}}")
print(f"Category: {{example['category']}}")
print(f"Language: {{example['language']}}")
if example['image']:
    example['image'].show()

Filtering Examples

# Filter by language
russian_attractions = train_data.filter(lambda x: x['language'] == 'ru')
english_attractions = train_data.filter(lambda x: x['language'] == 'en')

# Filter by category
museums = train_data.filter(lambda x: 'Museum' in x['category'] or 'Музей' in x['category'])
national_parks = train_data.filter(lambda x: 'National Park' in x['category'] or 'Национальный парк' in x['category'])

# Get only records with images
with_images = train_data.filter(lambda x: x['image'] is not None)

# Filter by location
tbilisi_attractions = train_data.filter(lambda x: 'Tbilisi' in x['location'] or 'Тбилиси' in x['location'])

Iterating Through the Dataset

# Process each attraction
for attraction in dataset['train']:
    name = attraction['name']
    category = attraction['category']
    description = attraction['description']
    
    # Process image if available
    if attraction['image'] is not None:
        image = attraction['image']
        # Your image processing code here

Attraction Categories

The dataset includes the following types of attractions:

  • National Parks: Protected natural areas with diverse ecosystems
  • Museums: Cultural, historical, and art museums
  • Fortresses: Medieval and ancient fortifications
  • Monasteries: Religious complexes and churches
  • Natural Attractions: Mountains, lakes, waterfalls, caves
  • Archaeological Sites: Ancient ruins and historical excavations
  • Architectural Monuments: Historic buildings and structures
  • Parks and Gardens: Urban and botanical parks
  • Viewpoints: Scenic overlooks and panoramic locations

Use Cases

Tourism and Travel

  • Building travel recommendation systems
  • Creating interactive tourist guides and mobile apps
  • Developing AR/VR tourism experiences
  • Route planning and itinerary generation

Computer Vision

  • Image classification and categorization
  • Object detection (identifying architectural features)
  • Image similarity search
  • Visual place recognition

Natural Language Processing

  • Multilingual text classification
  • Named entity recognition for locations
  • Cross-lingual information retrieval
  • Text summarization and generation
  • Translation quality assessment

Research Applications

  • Cultural heritage preservation studies
  • Geographic information systems (GIS)
  • Tourism pattern analysis
  • Multilingual dataset research

Image Licenses

All images are properly licensed and attributed. Common licenses include:

  • Creative Commons Attribution (CC-BY): Free to use with attribution
  • Creative Commons Attribution-ShareAlike (CC-BY-SA): Free to use with attribution and same license
  • Public Domain: Free to use without restrictions

Each image's specific license and author information is provided in the license and photo_author fields.

Data Collection and Processing

The dataset was compiled from publicly available sources with proper licensing. All images have been verified for appropriate licensing and attribution. Descriptions were collected in both Russian and English to support multilingual applications.

Quality Assurance

  • Verified image licenses and attributions
  • Cleaned and standardized location information
  • Removed duplicate entries
  • Validated data completeness and consistency

Limitations

  • Image Coverage: Approximately 11% of records do not have associated images
  • Image Quality: Varies based on source; resolution ranges from moderate to high
  • Description Length: Some descriptions are more detailed than others
  • Geographic Coverage: May not include every attraction in Georgia
  • Temporal Currency: Information reflects the state as of December 2024

Ethical Considerations

  • All images are used in accordance with their specified licenses
  • Proper attribution is provided for all photographers and content creators
  • The dataset respects intellectual property rights
  • No personal or sensitive information is included

Citation

If you use this dataset in your research or applications, please cite:

@dataset{{georgian_attractions_2024,
  title={{Georgian Attractions Dataset}},
  year={{2024}},
  publisher={{Hugging Face}},
  howpublished={{\url{{https://huggingface.co/datasets/{repo_id}}}}}
}}

Contact

For questions, issues, or suggestions regarding this dataset, please open an issue in the dataset repository.


Dataset Version: 1.0
Last Updated: December 2024
License: Dataset metadata is CC-BY-4.0; individual images retain their original licenses as specified