Skip to content
Colin Greenstreet edited this page Feb 1, 2017 · 6 revisions

Value Proposition:

  • Enable the creation of single image repository on the fly, sourced from multiple public and private repositories
  • Provide enhanced productivity in the creation of such on the fly repositories through the use of algorithmic and pattern driven searches of image metadata and images themselves
  • Provide date and period search functionality using confidence intervals around assigned dates and periods
  • Provide mapping functionality to tie images to rigorous geolocation, with confidence intervals placed on the geolocation point

Use case: Early Modern Thames imagery

  • #maphackathon Pinterest board [819 pins, as of 29/01/2017, 15:21]
  • At the February 11th-12th, 2017 hack, pull image data on Early Modern Thames theme from The National Archives (UK) and from the British Library as demonstration of the use case using multiple APIs

Inspiration

Proposed functionality & process flow

  • Images pulled from multiple repositories to cloud based picturemash repository via APIs, using automated recognition of content & geolocation, with community added precise geolocation, thematic tags, dates and date ranges

  • Build POC product for use case of Early Modern Thames imagery

Functional components

  • Ingestor for hand created Pinterest boards
  • Automated geolocation of hand selected pins in hand created Pinterest board
  • Automated tagging of hand selected pins [tagging by topic]
  • Automated dating and periodicisation of hand selected pins, using (a) pre-existing pin metadata (b) automatic (to be developed) tools [precise dates, if in pin metadata; data ranges, if in pin metadata; precise dates or date ranges using image recognitiona nd similarity algorithms]
  • Community addition or correction of pre-processed geolocation data per pin or image
  • Community addition or correction of tagging of pre-processed pins or images
  • Community affition or correction of dating and periodicisation of pre-processed pins or images