When Cowgirls Meet Computers
Interactive Installations at the National Cowgirl Museum & Hall of Fame
by Kevin Walker
2002
Introduction
Process
Poster Exhibit
Bronc Exhibit
On the Road
Arena Fashion
Annie Oakley Web Kiosk
Spirit Projections
Hall of Fame
This was my first major project working with West Office Exhibition Design, and it began in January of 2001. The museum opened in June 2002. The computerized exhibits ranged from a simple web kiosk to complex, interconnected systems, with the overall goal to simplify visitor and staff interactions with the computers. I enlisted Daniel Cummings to take on the more complex Director programming, and Director Xtras were provided by Daniel Rozin, Herbert Diamant, and Geoff Smith. BBI Engineering did the hardware installation, and Maltbie Associates did exhibit fabrication. Hired hand Ra Byn has been invaluable in installation and remote troubleshooting. And the entire museum staff was outstanding in their generosity, patience, and resourcefulness. The exhibitions were all about the cowgirl spirit, and they certainly embodied it. My wife was also the project manager, and without her, we'd all have been overwhelmed and directionless.
This type of museum was also new for me, having worked almost exclusively with science content. But the stories and the culture turned out to be amazing, and this comes through in the exhibits. After Sept. 11, the project seemed to generate additional interest and significance. The museum, especially on opening weekend, when it was filled with real cowboys and cowgirls -- including many Hall of Fame honorees -- has incredible emotional impact.
From the start, I was determined to bring the lessons I've learned over the years to bear on my process (which is summarized here). West Office also has refined a working process over the years, consisting of concept design, design development, final design, and installation phases. Early on, I created a simple web site which proved helpful for organization, communication, and review, between people in different parts of the country. West Office's FTP site was used extensively for sending things back and forth.
This exhibit runs on a Powermac G4, with buttons hooked through a BBI show controller via serial RS-232. A small JVC digital video camera feeds into the computer's Firewire port, and a flash also is controlled through the show controller, though for effect only. Everything takes place in Director, with live video compositing accomplished with Danny Rozin's VideoMask Xtra, still frame export with Herb Diamant's f3qtEdit Xtra, and receipt printing (complete with 1-bit thumbnail-sized poster) done with the Print-o-matic Xtra. The computer deposits the snapshots on an iMac in the shop via Appletalk over TCP/IP, where another Director program allows barcode scanning and simple searching through 30 days worth of images. It prints 5x7 images on an Epson 1280 Photo printer, also using the Print-o-matic Xtra.
Dan Cummings wrote the code and we began testing right away. We considered still cameras, or doing the compositing with special hardware (such as a Videonics mixer), but decided on the VideoMask Xtra because it gave us maximum flexibility and customizability, and we were familiar with Danny Rozin's excellent work. We tapped him later in the project to add DV capability and mirroring to his Xtra. We selected the $800 JVC GR_DVP3u camera for its manual settings and small size, though I later came across a Firewire webcam called the iBot, which would work just as well. VideoMask has a lot of parameters to tweak, but lighting of the greenscreen still proved important. (Originally a bluescreen, it was switched to green due to the prevalence of blue jeans in Texas.) The monitors were to be mounted vertically, so having everything turned sideways proved an additional challenge. And to make the monitor act as a mirror, we had to flip everything horizontally, then flip it back after the picture is snapped so that actual orientation is preserved and t-shirts don't appear backwards.
The Dymo printer also proved unstable under constant usage, with paper jams and USB conflicts. We ditched a (CAT 5) USB extender cable, and barely had enough room to run a direct, 14-foot USB cable after Maltbie quickly and seamlessly made a channel in the floor. Finally, we switched to a Ithaca POSJet, designed for more rugged retail use. The print quality is not as sharp as the Dymo, but the paper rolls are cheaper and easier to change. After we swapped the adapter and printer, most of our problems went away.
All the AV hardware sits in a crate on the floor, visible in the photo above.
The Epson printer has decent print quality, but takes at least 30 seconds to print, and during the packed opening weekend this caused up to a two-hour wait for printouts. A purportedly faster HP was hastily purchased, but did not prove much faster, and its print drivers were not as customizable; even with the Epson, it was hard to get the printout simply to be centered on the page, without some manual adjustment by the staff, nearly foiling our goal to go mouseless. But we went back to the Epson ultimately. It was inexpensive and thus easily replaced, but of course the museum goes through a lot of ink and paper.
Dan coded in some nice searching and error-checking features. The poster software sends an image every time it snaps one, and the shop software checks for new ones every 30 seconds. Barcode scanning a ticket immediately brings up the right poster, or if it can't find it, goes to a visual search screen. I stuck to a minimal screen design, and added some archiving capabilities -- every night the computer moves the day's images to an archive folder, deletes images more than 30 days old. Everything starts up and shuts down automatically.
Further information including technical details are here
This exhibit turned out to be the single most complex and challenging project I've ever done, and I've done some fairly complicated ones. But it turned out to be a fantastic experience for visitors -- perhaps the most popular exhibit in the museum.
The mission, as with the poster exhibit, was to create a fun hands-on activity which fully immerses the visitor in the cowgirl spirit. There are some fantastic rodeo cowgirls with incredible stories, and this also provided an opportunity to give some bronc-riding pointers from them, as well as to show some great old rodeo footage.
So far, so good. Now, here's where the tech comes in. The bronc is in front of a greenscreen, like the poster exhibit, and there is an identical video camera and computer crate. There is a 37" plasma monitor, again as in the poster, facing both rider and the entry area to the exhibit. Here is the complete process: You saddle up and press a button near the saddlehorn to record a 10-second ride, which is then edited into an old newsreel.
There were the same USB conflicts, and we switched printer and serial converter. To upload the movies, we originally tried the FTP Xtra but it proved unstable, so I switched to Applescript and added a step to run the movies through Quicktime Player to compress them. Jumping into Applescript cold, very late in the game, was an adventure, but it was something we had considered early on, and back at AMNH, Steve Godun had written some Applescripts for kiosk Macs, which I was able to use as a starting point. Applescript is fairly simple, especially after wrestling with Lingo, Director's scripting language, and in fact, I learned a ton of that too on this project, thanks to Dan. When his phase ended, we did a line-by-line walkthrough of all the code, knowing that I would have to update it later.
Working out the uploading process has taken about two months past opening day, and as I write this we seem to have finally nailed it down. Again, a learning experience, but ultimately rewarding. My next Mac-based project will be in OS X, and I have heard good things about its stability and Applescript capabilities. We didn't have major system-level problems with OS 9, but they took so long to work out that the Macs didn't appear to measure up to the PCs, which have been running mostly trouble-free since software was first installed on them, early on. But the PCs are running very simple programs and not dealing with video; if someone can show me how to do the above on a PC and keep it stable, I'd love to see it. I continue to maintain that Macs are better for public installations, with a few exceptions, as noted below.
Further information including technical details are here
Here is an interesting, similar project, done by a team in Hong Kong in 2003.
This one, in fact, was supposed to run on a Mac. It is a touchscreen program featuring rodeo stars telling some of those great old stories. Visitors sit on an old truck tailgate and put on headphones. We contracted the software to Tim Halloran, who did a wonderful job. He developed it on and for the Mac, and we purchased a 733MHz G4 for it to run on.
In this case, the computer was to be in a remotely located control room. But running the cabling between it and the screen required converting USB to serial, then back again, and neither the USB or serial touchscreen drivers liked this. We happened to have a spare PC in the control room (all the PCs were small, new Dells). We weren't sure if it would run properly, if at all, but after installing Quicktime, it has run nearly without a flaw since then. Nice work Tim! This also freed up the Mac for the emergency described above.
I wasn't able to test all the hardware ahead of time, but Ray sent me his program, and my simple Director program was finished upon my arrival on-site, so this was the first software installed. We were bedeviled for a while when it wouldn't owrk, but as often happens, this one had the simplest cause -- a loose wire.
Annie Oakley Web Kiosk
I had done a similar piece on the Mac at AMNH, of a fossil shot in HDTV. In this case, the stills were animated in After Effects, and encoded to a high-bitrate MPEG-2 file. For the previous project, I had used a WiresStram MPEG-2 card in a Mac G4, which could be controlled with a Director Xtra. The really nice thing that it did was allow you to put text and graphics on top of the MPEG stream, so you could set a resolution of 1280x1024, use the card to scale up the MPEG nicely while maintaining excellent quality, and overlay high-resolution, animated titles (or whatever) over top.
Unfortunately, neither the card nor Xtra is available anymore. Smokey Forester (his real name), who ran the Science Bulletins at AMNH, sold me on the $80 Hollywood Plus card for the PC, and sure enough, it scaled up just as nicely (having the same graphics chipset as the Wired card). But the MPEG-2 Xtras for the PC don't allow laying anything over the MPEG stream. No matter; for this piece I burned the titles into the MPEG, and dispensed with Director altogether, running it in Windows Media Player.
The screen was originally translucent all around the projection, making the projection look like it was floating in a big black hole, not filling enough space. Bill suggested painting the interior of the frame white around the projection, and that gave us the nice effect, discernable here, of the image floating just in front of the wall.
I was not able to test any of this beforehand, and these exhibits were some of the last to be installed. The three sensors could not be placed close enough together for their beams to overlap, so one problem was that it would send multiple readings from a finger drifting between them. (A whole hand worked fine, but was not intuitive.) I solved this with by taking the first reading then blocking others for a second or so.
Another problem with just one axis of sensors is that touching anywhere in the vertical band blocks the beam. I tried to play this up in the interface design, working with Henni Yama at West Office, by making the buttons very large and visible, and exposing the inner workings by using some virtual up-lights to subtly indicate vertical beams.
There are over 150 honorees in the Hall of Fame, and the simplest and most intuitive way to search them is alphabetically. With only 11 touchable areas, we first tried making scrollable columns of names -- you touched and held down on a button, and released when the name you want was highlighted. But the multiple readings foiled this, and it was just too confusing. After working with the sensors a bit, I changed it so that you simply choose a letter of the alphabet, and it goes to the first name beginning with that letter. Once you are on an honoree's bio screen, you can navigate in either direction through the names.
New honorees are added every year. As with the clothing rack interactive, this program was designed to be easily updated by the museum, by replacing files in a folder, and updating a text file listing the names. I had considered making it updatable via web browser. This would be simple to implement, but doesn't really make sense, since someone still should check the program in the exhibit.
We wisely decided to tackle the most complex items first, and this one came first. Visitors create their own mock movie posters by selecting a poster design from a button on the wall, then standing in front of a greenscreen. A monitor facing the visitor acts as a mirror, allowing them to position themselves against the background; a monitor facing outward shows the same thing. There are five poster designs, each designed to appeal to single or multiple, male or female, visitors. After a countdown (visual and audio), the picture is snapped and a barcoded receipt is printed, which can be taken to the museum shop to purchase a printout.
With a lot of interconnected hardware and processes, we tried to simulate the exhibit conditions and test all of the pieces independently and together. West Office has a nice, big former factory building and a full shop, ably run by Jim McGee, and he put together a nice greenscreen setup.
The G4 and show controller proved perfectly able. In our mockup we used a Keyspan adapter to convert serial to USB, but we encountered problems between it, Director, and the USB printer, and on-site we switched to a Stealth Serial Adapter, (long ago recommended to me by Mac guru Steve Godun).
Here are some test composites, one of many dour and poorly-lit shots of me staring at the computer screen, and a test thumbnail print. The initial poster designs I did here were later made beautiful by April Banks at West Office.
It was Bill's idea to distill the instructions down to three simple steps. Testing with the mockup caused us to switch steps 2 and 3. Pressing a button in step 1 starts the countdown, and having the visitor take the ticket first caused them to spend more time at the instructions. Plus, the ticket would appear in the picture. Now they push a button, go immediately to the screen, then pick up their receipt on the way out. Because space was fairly tight, we removed a separate exit, so the visitor enters and exits through the same gate.
Meanwhile, down in the shop, the iMac was chosen for its small footprint, and our goal was to avoid a mouse and keyboard to save additional counter space, confining ourselves to a barcode scanner and keypad (which I velcroed to the iMac). The iMac, thankfully, has one more USB port (3 total) than the G4. It is a sharp visual contrast to the cash register PC, with its cash drawer and CRT monitor. We did not attempt to link the two systems.
So we decided to put visitors into the rodeo arena, and into the old newsreel footage. Safety issues, of course, immediately came to the fore, and we decided against an electrical bull like those in bars. We found Bullmaster, which is like a full-sized version of the playground spring-based ride, and is used for rodeo training. Maltbie sculpted this beautiful horse to go onto the base, cowgirls added real saddle and bridle, and the whole thing was set onto 10 inches of blue padding, making it sort of look like a swimming horse.
In this case, the greenscreen removal is not done on the live video feed, but processed afterward. Compositing, editing into a new clip, and saving a Quicktime movie takes about 30 seconds on the 933MHz G4 -- a long time if you're concerned with maximum visitor throughput. We use that 30 seconds to tell the visitor to go to the museum's web site, where they can download the movie. The completed movie plays back, and again, a ticket is printed, this time with login and password, so that only the ticket holder can view the ride. Every night, the bronc compresses and uploads the movies to the web server. The web component was added fairly late in the process; originally, rides were to play back on a second monitor, but it was eliminated. The instructional footage was also to run on the second monitor, but we were not able to secure the rights in time. Some printed tips appear on a graphic panel at the entrance.
Quicktime has all of the editing capabilities built in, and Herb Diamant modified his f3 Xtra so we could add tracks and edit clips together . (Compositing is done with a Quicktime plug-in from Buena Software.) Here is a test movie, taken at West Office with our makeshift bronc; and exhibit designer Steve Tornallyay in the final exhibit.
There were some difficulties with the exhibitry too. We had used the same full-size mockup as for the poster exhibit, and this served us well. But due to delays, we were not able to install and test fully until just days before the opening. One piece of hardware we were not able to test ahead of time was the transmitter connecting the button to the show controller. To avoid running wires through the flexible shaft in the center of the bronc, a wireless method was chosen -- literally, a garage door opener (pictured, upper right). It eventually worked fine, but as it was quickly being hooked up the night before opening, it apparently sent 110 Volts over its data connection, instantly destroying, in order, the show controller, serial converter, computer, and camera. The software was installed and working, and when I went to boot up for opening, it was odd to find nothing working. We were extremely fortunate to find a spare camera at the local Cirtuit City, and to have on-hand a spare show controller and Mac, and I was acutally able to have it running about halfway through the members' preview.
A moving clothes rack is controlled by a touchscreen, stylishly encased in wood and red leather. It was a good idea and it's very effective. The software is simple: a static page for each outfit, with zoom-in buttons to particular details. The outfits were photographed by the museum staff and the program is updated by them by swapping image files in a folder -- easier than changing the actual outfit.
This runs on a PC, and to control the rack, the computer simply sends a character out the serial port (from Director via Geoff Smith's Xtra, which was used for serial communication on all of these projects). Ray Golley at Maltbie wrote a simple program for the Basic Stamp microcontroller to use the letter into trigger a set of relays, which simulate button presses on a standard dry-cleaning rack controller.
Rounding out the Arena gallery, this is a simple web kiosk, the screen this time encased, with trackball and button for access. Annie Oakley film footage runs as an attract loop. Dave Heyliger's Web Kiosk Commander software is the only reason to use a PC instead of a Mac in web kiosks; to combat the inherent instability of browsers and Windows operating systems, he has put in some nice auto-rebooting features. But a stable Mac-based web kiosk is on the horizon, in the shape of an iMac (or eMac) running OS X.
The other reason to use a PC over a Mac is for inexpensive, high-resolution video playback. This is a linear video of gently moving stills overlaid with words representing the cowgirl spirit, and there are two variations, projected in the rotunda outside the exhibits. The words are also etched into glass on the surrounding walls, framing small artifact cases.
To round out our backward tour... As you enter the museum, the first floor rotunda is the Cowgirl Hall of Fame proper, and every honoree has a star on the wall. To enable visitors to search for a name or learn more, we created this giant touchscreen.
Giant touchscreens are nothing new, and there are lots of ways to make a projection or a large monitor interactive. In this case, we used infrared sensors, and when a beam is broken, a BBI show controller sends a number to the PC. There were some technical constraints: To minimize the size of the frame, we only put sensors on the top and bottom; and to match the size of the real stars on the wall, we had to use three emitters for each hot spot (these are visible on the bottom of the frame, and a single receiver at the top captures all three).