pastvu / README.md
nyuuzyou's picture
Super-squash branch 'main' using huggingface_hub
5b2f548 verified
metadata
pretty_name: PastVu Historical Photographs
size_categories:
  - 1M<n<10M
task_categories:
  - image-classification
  - image-to-text
  - text-to-image
  - object-detection
  - zero-shot-image-classification
annotations_creators:
  - found
language:
  - ru
  - en
multilingality:
  - multilingual
source_datasets:
  - original
configs:
  - config_name: default
    data_files:
      - split: train
        path: pastvu_*.tar
    default: true
tags:
  - image
license: other

Dataset Card for PastVu Historical Photographs

Dataset Summary

This dataset contains approximately 2,093,000 historical photographs from PastVu.com, spanning the years 1826-2000. PastVu is a collaborative project that allows users to view historical photographs on an interactive map, providing a unique temporal and geographical perspective on historical documentation. The dataset includes images downloaded at the best available resolution and is structured in webdataset format for efficient processing. The collection covers a wide range of historical subjects including architecture, street scenes, cultural events, portraits, and everyday life, primarily focused on Russia and former Soviet territories, though it includes photographs from around the world.

Languages

The dataset is multilingual:

  • Russian (ru): Primary language for titles, descriptions, and metadata
  • English (en): Secondary language with some translated content and region names

Dataset Structure

Data Fields

Each photograph in the dataset is accompanied by comprehensive JSON metadata containing the following fields:

  • type: Integer representing the content type
  • adate: ISO timestamp of when the photo was added to PastVu
  • author: String containing the photographer's name
  • cid: Unique content identifier
  • dir: String indicating image orientation/direction
  • file: Original URL of the image file
  • geo: Array containing latitude and longitude coordinates [lat, lon]
  • h: Original image height in pixels
  • hs: Scaled image height in pixels
  • ldate: Last modification timestamp
  • title: Photograph title/description in original language
  • user: Object containing uploader information:
    • login: Username of the uploader
    • avatar: URL to user's avatar image
    • disp: Display name
    • ranks: Array of user rank badges
    • sex: Gender identifier
  • w: Original image width in pixels
  • ws: Scaled image width in pixels
  • year: Primary year the photograph was taken
  • year2: End year (for photographs spanning multiple years)
  • s: Status or quality indicator
  • y: String representation of the year
  • waterh: Watermark height
  • waterhs: Scaled watermark height
  • watersignText: Watermark text content
  • watersignTextApplied: Timestamp when watermark was applied
  • r2d: Array containing additional geographical reference data
  • frags: Array of image fragments or regions of interest
  • regions: Array of geographical region objects containing:
    • cid: Region identifier
    • title_en: English region name
    • title_ru: Russian region name
    • phc: Photo count in region
    • pac: Active photo count
    • cc: Total content count

Data Splits

The dataset contains a single split:

Split Number of Examples
train ~2,093,000

Total dataset size: Approximately 2,093,000 entries

Data Format

  • Images are stored in webdataset format across 2,094 tar files
  • Tar files are named sequentially: pastvu_000000.tar to pastvu_002093.tar
  • Approximately 1,000 images per tar file
  • Images are downloaded at the best available resolution from the original source
  • Each photograph has corresponding JSON metadata
  • Complete metadata is also provided in pastvu.jsonl.zst
  • Geographical coordinates are provided for most images, enabling spatial analysis

Temporal Coverage

The dataset spans 174 years of photography:

  • Start year: 1826
  • End year: 2000
  • Peak periods: Soviet era (1920s-1990s) with extensive documentation

License Information

Licensing Structure

The licensing for this dataset follows the original PastVu.com terms and conditions. Users should be aware that:

  • Images may have varying copyright statuses depending on their age and origin
  • Many historical photographs may be in the public domain due to age
  • Some images may still be under copyright protection
  • Attribution to original photographers and PastVu.com is recommended
  • Any use of materials published by the user for commercial purposes is only possible with the permission of the copyright holder of the specific image.