While taking a look at the rear-projection system in one of the fancy rooms underneath the library here at RIT, we stumbled across this wonderful DIY how-to on how to build a rear-projection screen in your own bedroom. Using a shower curtain.

We were so inspired that the following Saturday we set up the projector and screens again, except this time we came prepared with (dun dun DUN) materials.

The screens in the special rear-projection room under the library are made of acrylics (looks like thick, frosted plexiglass), and at first we thought that that might be a viable option for us as well. However, a trip to Home Depot very quickly proved us wrong – first of all, plexiglass is very expensive in large quantities; second, its fragile, heavy, and hard to move, especially for such a large size. Finally, we were so inspired by the shower curtain screen that we had to try it ourselves.

We gathered an array of materials in the shape of 2 shower curtains (one translucent, the other white), 5 different fabric samples ranging in thickness and stretchiness, and a large stretchy drape cloth usually used for photo shoots. Over the next several hours we tested and retested these materials for quality of projection. As expected, the stretchy ones worked the best – first of all, they are thinner and let the most light through, which allows for the best picture quality. Secondly, they do not wrinkle, which means we don’t have to iron huge sheets of fabric before every set up. Score!

The more we thought about the setup of the screens, the more we realized it needed to change from our original mockup. The biggest issue came when we read the guidelines for displaying installations in the fieldhouse, which gets the most traffic at Imagine RIT: no tents or roofed constructions, or flammable materials. Having some sort of roof over our heads for the installation is going to be pretty important – we will need it to block out the extra light that will interfere with the silhouettes, what’s being seen on the projection screen, and with what the camera picks up. Taking this into consideration, we decided to split the installation into two roofed sections so that the light is blocked out from directly above the players (so it does not interfere with the silhouettes), and also so that the light is blocked from directly above the screen, so that the visibility is not affected. In the dead space in the middle will be a wide walkway for passersby to be able to look in and see what’s going on inside. This will also help with the fire hazard regulations. I adjusted the mock up.

The screen that filters the light behind the people is curved in this setup to accomodate for it being smaller and still taking up the full view of the camera. In the original setup, the screen would have had to be about 16 x 18 feet in order to cover the whole background of the field of view (FOV) of the camera. This way, if we curve the back and lean it over the FOV, the same effect is achieved with more structural integrity and less heavy things to move.

The only problem with this setup is the amount of room it requires – that’s about 35 feet in length alone. Actually, backlighting the entire installation almost doubles its length, since the projector has to be about 12.5 feet away from the screen in order to project an image that’s about 7 – 8feet tall. We want this to be a fully immersive experience, so the projection needs to be very large, so the deadspace in front of the projector has to be pretty large as well. We could always try to save space by front-lighting the whole thing, which would increase the quality of the image too, but this presents the problem of having to hide the projector somehow – either constructing a structure for it above or below the screen.

Before going too forward with the project, we wanted to make sure that we tested our proposed setup to ascertain the dimensions, playing space, projectors, and cameras for everything we need. We used FluidTunes to test the camera and its sensitivity to light, dimension, and angle to the users.

We set up a camera on the floor at a 13 degree angle looking up at the players, and rigged up a sheet on some poles to provide a solid background for the program to recognize the silhouettes of the people moving about in front of the camera. This setup actually worked fairly well for interacting with FluidTunes, especially after we brought in a lighting kit to light the sheet from the back and create more of a contrasted silhouette.

We also tested a regular video camera for the feed in, which worked just as well as Linzi’s external iSight. Most likely the best solution is going to be using a lower quality camera, like the iSight, since the program will only need to detect dark pixels vs. light, and the actual image of the person will not need to be projected on the screen in high quality. By sacrificing quality no one is ever going to see we are actually optimizing the processing for the program itself, since the code will not have as many detailed pixels to rip through.

This first setup was mostly used to determine the sizes of the area that will be visible on screen and therefore playable. Linzi created this mock up to summarize our calculations.

Since we’re planning on setting up the display at Imagine RIT, we need to somehow keep the projector hidden from view of the visitors to the exhibit. While we could potentially construct an overhead berth for the projector, I think we all feel a little better about back-lighting the screen and keeping the projector behind it.