The project was to create a massive rear projection in the window of the gallery at 1506 Dundas St. West in Toronto.
The projection would be a massive mirror image of the street itself. Walk past the gallery, and see an image of yourself blown up across the entire window-front.
Wave your hand, and a screenshot of your image would be uploaded to an image blog online, parked at www.wave1506.com. You can go there now and see all of the images captured.
How we done did it:
The first step was to find a material to change the window into a surface that would allow for rear projection.
3M makes a material for this, called Vikuiti. We determined that it would cost over $6000 to cover the entire window in it, so that was completely out of the budget for our project.
We instead looked into some DIY methods for creating rear projection film. It occurred to me that the stick on “fake frosted glass” that you can get at Home Depot might do the trick. From our initial testing, it turned out to work just as well as the 3M material, and would only cost us about $150.
A very wide angle, bright projector was used to project the image. It was important that it was HD for the size of the projection, and also very bright due to the fact that the street is not very dark, being lit from streetlamps. We infact did not want the street to be dark, as then our camera would not capture any light, but that meant a very bright projector was crucial to our project.
In order to capture the image of the street and rear project it back, we needed to hide a camera inside the projection itself, but the camera needed to be 1080P HD (because the size of the image we were projecting), and work well in low light, but not be very big or else it would obstruct the projection.
We settled on Logitechs highest-end webcam. It worked well mounting it to the beam in the middle of the window, and running the wires down the beam so they didn’t obstruct the projection.
A Kinect Controller was used to capture peoples movements, hidden at the base of the projection.
We used the node-based program VVVV to handle all of the image processing and triggering. A framework created for the Kinect called Kinecursor was used to recognize peoples hand gestures in the form of a wave, and input that into VVVV. Whenever someone would wave their hand, Kinecursor would sense this and send data to VVVV, triggering a screenshot to be taken of the projection video.
These photos would dump into a folder on the computer (a Mac Mini running a specially hacked, uber minimal version of Windows 7). An apache server right on the Mac ran the website at http://www.wave1506.com. Everytime a photo was dumped into the folder, a PHP script would add it to the custom made photo blog that we created. The blogiste is a responsive site; it was very important that it would work on mobile phones as people would want to check their image right infront of the projection. A timestamp is added to the photo so it has a sense of time and place. Tthe website is just a simple single page of photos that scrolls for an eternity, which would have seriously overloaded the users browser and our own server if it loaded all the images at once everytime the page was refreshed. JQuery was used so that only the images within the users viewport on their browser would load, in so that all 1000 or so photos would not load everytime you went to the website.
At the front of the window were directions on how to use the installation, as well as the website where you could view your images. An NFC sticker allowed users with NFC enabled smartphones to tap their phone against the directions and be automatically taken to the website.
Here are some pics of the entire rig.