About this project
What is Aerial Bold?
Aerial Bold is the first map and typeface of the earth. The project is literally about "reading" the earth for letterforms, or alphabet shapes, "written" into the topology of buildings, roads, rivers, trees, and lakes. To do this, we will traverse the entire planet's worth of satellite imagery and develop the tools and methods necessary to map these features hiding in plain sight.
The entire letterform database will be made available as a “usable” dataset for any of your art/design/science/textual projects and selected letterforms will be made into a truetype/opentype font format that can be imported to your favorite word processor.
Only with your support can we turn topography into typography and bring Aerial Bold to you!
Contrary to popular belief, much of the world has not been mapped. While satellites orbit around the earth taking thousands of images each day, we have limited idea about what unique features actually live on those photographs.
Aerial Bold is as much about developing new methods of mapping features on the earth’s surface, as it is about generating the first map and typeface of the planet. It is our intention to offer non-domain experts (e.g. artists, designers, citizen scientists, etc.) a set of tools to source their own datasets and inspire people from all backgrounds to explore geographic data. The importance in kickstarting Aerial Bold is to showcase the creative and technical process of “making your own data”.
As a two-man army, we will be experimenting with “new” ways to train image processing algorithms to find geometric patterns and use these scalable methods to find the shapes of the alphabet over the entire earth.
On a small scale, such as in a community or small city, image processing tasks don’t pose a big challenge, however on a large scale (e.g. the entire earth, or the entire US), the computational requirements can be paralyzing. This is why we’re asking for your help.
With your support we will be able to purchase dedicated computing/server space to run our satellite image based alphabet searching algorithm. The funding support will help keep the typeface, the maps, and the alphabet metadata (e.g. geo-coordinates, address, etc) alive in cyberspace for people to download, use, view, and explore.
Who Are We?
Benedikt is a computational and speculative designer from Germany with over 12 years of experience in design, digital media, and computer programming. He is a co-author of the best-selling book on Generative Design, has worked as a research assistant at the Massachusetts Institute of Technology (MIT), and has a MA from the Royal College of Art in London. His projects have been featured in Wired, the LA Times, the New York Times, Ars Electronica, the Japan Media Arts Festival, and a number of other blogs and publications. He is co-author of the Big Atlas of LA Pools and lecturer at HfG Schwäbisch Gmünd Institute for Applied Science. In his free time, Benedikt escapes the city to the Austrian Alps and enjoys a mild darjeeling tea.
Joey is a geographer from San Francisco, California, USA working in geospatial technology, data visualization, and digital media. He has a BA from UCLA and has worked as a research assistant at the Massachusetts Institute of Technology (MIT). He is actively involved in data visualization and science communication projects and conducts research on smart cities and urban climate science. He is a co-author of the Big Atlas of LA Pools and is currently pursuing a MSc at the University of British Columbia in Vancouver, Canada. When Joey is not making maps, he can be found on long bike rides, skateboarding, or enjoying the latest This American Life.
Pledges & Rewards
Example ABC Dataset
How does this all work?
The processing pipeline graphic shows our planned method for deriving the letterforms.The checked, green circles indicate the tasks that we have already completed and/or prototyped and the unchecked, yellow circles indicate the tasks that rely on your funding support for processing power.
The process we are developing is almost completely automated, with more manual steps required for quality checks and input training data. We've mapped out our methodology, we're working on up-scaling and training the algorithms which is no small task.
Example Images of Processing Pipeline
This is a sample of an aerial image that shows some buildings in the shape of the alphabet letter "A".
The image below shows the results of the "image segmentation" process. Image segmentation simplifies an image such that it is easier to differentiate between features. With a segmented image, its becomes more simple for the computer to recognize text.
The image below shows the buildings (in red) and natural land category (in green) vectors from openstreetmap. The alphabet letter "A" shapes can be seen clearly demarcated.
The image below shows the vector data overlaying the aerial image. This combined dataset will help us to find Aerial Bold on the earth's landscape!
Kickstarter Video Credits
Risks and challenges
This is an ambitious project that will require experimentation and a large amount of data wrangling. While we are dedicated to deriving earth’s letterforms by developing the best possible methods using computer vision, there will most certainly be some cases in which more manual labor will be employed. This will be especially true for regions of the world where imagery and data are of lower resolution and less available and when validating each letterform (e.g. is what the computer thinks is an alphabet shape indeed an alphabet shape?).
Through brute-force approach (manually finding letters), we’ve already compiled a complete alphabet dataset, and in many cases have found multiples of each letter. We know there are more out there and we want you to have access to all of them.
Deriving data from satellite imagery is and can be a tricky thing from a legal point of view. All of the companies that provide aerial imagery online (e.g. Google, etc) have specific terms and conditions, which almost always forbid any data mining of the imagery. For some aerial imagery providers, however, there are provisions that allow you to derive data from their imagery (e.g. if the data is contributed to OSM, openstreetmap.org).
These are indeed important considerations, but we’re optimistic and have already planned out ways forward. First, we approach one of the aerial imagery providers and they simply allow us to use the data for this project (we’ll keep you updated!). If we have no luck in accessing pre-made global imagery tileset, no problem! We can start with the parts of the world where imagery is available at high resolutions under public domain (e.g. in the United States) and continue from there. Regardless of the imagery source, we will combine the imagery data with OSM data which includes quite a comprehensive global database of vector buildings, roads, trees, lakes, rivers, etc. Together, these will serve as the basis of our alphabet shapes.
None of these approaches are mutually exclusive.
We are confident that we can bring Aerial Bold to you!Learn about accountability on Kickstarter
By pledging $90, you get to own a letter. This is how it works.
After we have completed the project, we will have a extensive database of all the letters from A-Z. There will likely be hundreds (dare we say thousands?) of letters from all over the world, many of them repeating.
You will then be able to pick a letter of your choice (e.g. "an A in Berlin", or a "K in Salt Lake City"). Along with the letter, we would invite you to send us a short text that would be written into the metadata of the letter image. An example could be, "G: For all the geographers out there".
Since in the end, the dataset will go "public" or be "opensourced", we would let you decide if you would like it to be excluded from the opensource dataset or included. If it is excluded, that letter is yours and will not be available to the others. If not, there is the chance that the letter you chose can be incorporated into other projects. This could be interesting in a way, because then your letter and your short text might "live on" through those other projects.
We hope this is understandable and agreeable. Thanks so much for your support and being an integral part of making this happen!
What do the tools/code/algorithms package ($70) look like? Is it a cloud-based tool? Is it a tool packaged in an .EXE or .DMG?
The tools/code/algorithms package will take the form of a code repository (e.g. Github) with instructions, documentation, and examples about how to set up the processing pipeline on your own personal machine. But it will be less a desktop app and more something like command line tool.
The main programming language and associated libraries are currently based on Python (python.org). Since Python and the libraries we are using are all opensource, accessing the base libraries are simply a matter of downloading them from your favorite source. With our bundled package of tools/code/algorithms, the documentation will guide you through the setup and get you up and running.
Over the course of the project parameters mights change, hence it might be in the end a different programming language. But as we have already good results with python we are optimistic to stick with it.
Are the image processing tools/code/algorithms limited to satellite imagery or can they be applied to any random image?
The processing pipeline we are developing are focused specifically on finding features in satellite imagery tiles, like those we see on Google Maps, Bing, or Mapbox in the standard "image tiles" format ( ".../zoom/x/y.png"; see http://wiki.openstreetmap.org/wiki/Slippy_map_tilenames). Thus the pipeline will be optimized for analyzing these types of imagery.
However, this is not to say that the tools/code/algorithms can't be recycled and reworked for analyzing your own input imagery.
Support this project
- (30 days)