Public Art
| Infocard
| ||||||||||||||||
| ||||||||||||||||
Find nearby street art now!
|
| Kb |
Contents
Description by the Publisher
I made an app that pulled Instagram graffiti images, based on the #graffiti tag, and made it accessible via an iphone app. The app let you check which of the indexed images were closest, and surfaced the route to discover art nearby.
Public Art was an iOS application that made it possible to discover over 100,000 new street art and graffiti images from Instagram, a month.
The project started after I had created a photo pipeline from Instagram, where I indexed all images with lat/lon data and a series of hashtags. Based on this 'feature', I realized there was no effective way to browse the images. The result was an app that passed the device's own current location, and processed a search query for the nearest indexed images.
The project ran for 2 years until the 'features' used to pull Instagram images couldn't confidently return graffiti and street art. The cost of processing every indexed image to prevent undesired images was too much at the time.
Games to help label anonymous artists
Think Capcha for graffiti. I created a simple game for 'crowdsourcing' artist labeling. The game gives points for correctly labeled artists, while monitoring the community to see if consensus was met around the artist.
What's new about this?
I have not seen any global foundations that systematically preserve street art. I'm still figuring this out, but I feel like the current state of machine learning/CV + mobile devices + widely available internet makes this uniquely solvable problem.
Based on the wide spread use of social media, a larger amount of graffiti and street art is documented with multiple dimensions of valuable data (time/date/location). Using some basic forms of user generated image submissions, alongside crawlers that monitor certain social media hashtags, I know there is well over 100,000 new graffiti images that can be accessed and preserved a month. The number is likely closer to 400,000 new images.
Limitations in the Downloadable Version
No limitations
Product Identity
Unique Product ID: PID-7800230FD599
Unique Publisher ID: BID-580071109599
