What's this all about?
For years The New York Public Library has collected maps documenting our ever-changing metropolis. Originally commissioned by insurance companies to assess property value and fire risk, these street atlases contain a wealth of detailed information about a city now largely lost to us: buildings long ago destroyed, streets renamed, whole neighborhoods redrawn or redefined. Making these lost places findable via contemporary digital maps allows us to drill down through the layers of urban change and study the city in profound new ways. But harvesting the data isn't easy.
That's where the Building Inspector (and you) come in. This app is the latest in a series of public-facing tools designed by The Library to extract, correct and analyze data from historical maps. We're training computers to do the heavy lifting, and then distributing the remaining quality control tasks to smart, motivated citizens. The goal? To produce a comprehensive directory of old New York (or, as we like to think of it, a time machine).
With that information organized and searchable, we can ask new kinds of questions about history. It will allow our interfaces to drop pins accurately on digital maps when you search for a forgotten place. It will allow you to explore a city's past on foot with your mobile device, 'checking in' to ghostly establishments (Empire State Building, meet Waldorf-Astoria Hotel). And it will allow us to link other historical documents to those places: archival records, old newspapers, business directories, photographs, restaurant menus, theater playbills etc., opening up new ways to research, learn, and discover the past.
How Your Contributions Help
Currently, there are four different types of inspections you can perform. Each is designed to extract a particular type of data from the maps. Here's what the different inspections tell us:
- Check Footprints - This establishes a baseline (literally!). We show you the buildings the computer identified, one outline at a time. You tell us whether it is right, wrong, or close but in need of fixing.
- Fix Footprints - Take those slightly imperfect footprints identified by your fellow inspectors and get 'em into shape to be recorded for history.
- Enter Addresses - Getting those original street numbers will help us to reference specific buildings in their historical context (and, eventually, to see who lived/worked there).
- Classify Colors - The original mapmakers color-coded the buildings to indicate construction materials and use types (residential vs. commercial). Identifying the colors helps us index these important details.
How Is Accuracy Ensured?
Every time you inspect a building, you’re essentially casting a vote alongside your fellow Inspectors. We show the same footprint and task to several people and tally up those votes to decide whether they agree.
While each task has a different means to come to consensus, it’s useful to see an example. For Footprint Inspection, it’s coming to consensus on whether the computer-detected footprint is a Yes, No, or a Fix. We show the same footprint to at least 3 different people, and every 10 minutes we tally up the votes. If 75% or more agree, that footprint has reached “consensus” and the system removes it from the inspection queue. If the jury's still out, we keep the footprint in circulation until consensus is reached, focusing our collective efforts on the buildings most in need.
The main takeaway here is: relax! Your fellow Inspectors have got your back. It's no big deal if make a mistake once in a while, or make your best guess on an ambiguous case. The community will double, triple, and quadruple check you, and you them. Go Team!
How Did We Get Here?
It all starts with atlases
NYPL’s Lionel Pincus & Princess Firyal Map Division houses about half a million maps and 20,000 books and atlases published over the last five hundred years…
Our atlases span the globe across time and at many scales. You can trace the Dutch empire's global expansion through 17th Century world maps and pilot books; follow the battles of the American Revolution through 18th Century English atlases; and (as in this project) track changes to New York's urban landscape through maps commissioned by 19th Century insurance companies.
First we create a high-resolution digital image of the original paper map. We make copies of the image and store them for safekeeping in our digital repository. The maps are then formatted for deep zooming and are presented to the public via our Digital Collections and Map Warper websites, where they can be downloaded in high resolution and queued up for further transformation.
While images of old maps are nice, what our users really want is to compare maps from different periods. The trouble is, mapmakers rarely laid things out in ways that allow us to conveniently overlay the images. Locations appear in different parts of the page from year to year. To make matters worse, cartographers used different scales, pagination schemas, and (gasp!) sometimes didn't even put north at the top of the page. So we modified open source software and built the Map Warper, a tool suite for library staff and public volunteers to align maps to precise latitude-longitude points on the planet Earth. This normalizes all of our map images to the same digital base map, making comparisons easy.
Next up is identifying specific features from the maps. This involves identifying shapes (most commonly, building footprints) and converting them into geospatial vector data. This is fancy talk for a different way of handling images on computers. It treats them not as grids of pixels (called raster graphics), but as geometry (points and lines). Vector data is lighter weight and therefore faster to manipulate on a computer (think moving around a large canvas, zooming swiftly in and out). Staff and volunteers have been working to trace building shapes by hand but the work is time-consuming (and, frankly, pretty dull). So to get through these mountains of data, we're training computers to do a first pass on the maps, 'automagically' recognizing building footprints and other features. It works pretty darn well, but as with most image processing techniques, it's not perfect.
You take it from here! With the help of your fellow Inspectors, you’ll be checking the computer's work, and filling in other valuable details — unlocking the future of New York City’s past!
When we've processed all of our maps in this way, we'll end up with a vital digital record of the city's layout through the ages. You can learn more about the data and how to access it on our Data page. All inspection data produced by this project is released under a Creative Commons-0 Designation, meaning you can use it however you like.
The code for this project is also available in GitHub for you to use how you like.
The Lionel Pincus and Princess Firyal Map Division, NYPL
Geospatial Librarian - Matt Knutzen
Geodata Specialist - Mishka Vance
Lead Developer/Designer - Mauricio Giraldo
Director, Digital Library + Labs - Ben Vershbow
Product Manager - David Riordan
Historical Geospatial Fellow - Connor Gaudet
Additional Development Support - Paul Beaudoin, Brian Foo, and Matt Miller
Tony Marx - President & CEO, NYPL
Mary Lee Kennedy - Chief Library Officer, NYPL
Ann Thornton - Mellon Director, NYPL
Theresa Myrhol - Director, Library Services, Stephen A. Schwarzman Building, NYPL
Victoria Steele - Brooke Russell Astor Director of Collections Strategy, NYPL
Amy Geduldig - Manager, Media Relations, NYPL
Nora Lyons - Press Representative, NYPL
Jane Aboyoun - Chief Technology Officer, NYPL
Jeff Roth - Vice President of Strategy, NYPL
Micah May - Director of Strategic Planning, NYPL
Jennifer Anderson - Senior UX Designer, NYPL