A Clear Vision
Issue: Volume: 32 Issue: 2 (Feb. 2009)

A Clear Vision

At least one reason we make hand gestures when we talk is to help people see what we mean. Visualization is a critical part of understanding. On the computer, 3D visualization opens up the door to collaboration and decision-making. Not many people make decisions based on hand waving.
The field of visualization and, especially, immersive visualization is utterly booming, with changes happening in every area of the industry as designers and scientists push the technology for greater realism, while hardware architectures rise to the challenge. Systems are being built on what can only be called a grand scale. For instance, several visualization centers have been built that take advantage of Sony’s 4k projectors for super high resolutions and crazy realism—or, at least, it used to be crazy; now it’s just expensive.

What’s changing right in front of our eyes is the range of tools available for this purpose. A lot of the high-end visualization systems we have now came from work done in the 1980s by Carolina Cruz-Neira, Tom DeFanti, and Dan Sandin at the University of Illinois, Chicago, where the first CAVE (Cave Automatic Virtual Environment) was built. A CAVE can have three, four, five, or six walls, and stereoscopic views to give users the illusion of immersion.

Today’s systems, though, have evolved to suit a given situation. There are still CAVEs, and their numbers are growing, but there are also power walls and high-end workstations with 3D glasses, or auto-stereo­scopic screens that don’t require glasses, all providing people with a clear view into the abstract. Whatever the medium, the demand is for as much reality as possible—and in real time.

Europe Versus the US
Over the past decade or so, European development in 3D visualization has been on the rise. Jeroen Snepvangers of RTT (Realtime Technology), with headquarters in both Europe and the US, has seen both sides of the visualization/realism coin. He contends that the use of 3D visualization has increased much more in Europe than it has in the US, and a contributing factor may be a difference in attitude. Snepvangers heads RTT’s operations in the US, where, he says, the company has found considerably more interest in the creation of photorealistic images. In Europe, however, RTT is being asked to build visualization centers. “Everyone has to have a CAVE,” he adds.

Research in raytracing, graphics algorithms that track the path of light as it travels and bounces from source to objects, has been very active in Europe, including at Fraunhofer Institute in Germany and at universities in Lund and Gothenburg, Sweden. “Even five years ago, CAVEs were more popular in Europe than in the US,” says Snepvangers.

But, there have been considerable advances. Five years ago, Snepvangers says, the interactivity worked well in immersive applications, but the imagery didn’t look photoreal. That’s where RTT Europe has put a lot of its efforts. “Reality is on the priority list of our clients,” he notes.
HP, in conjunction with DreamWorks, developed the Halo system so executives can collaborate without technology getting in the way. With zero latency and high-definition resolution, participants soon forget they’re not in the same room with one another.

The same sentiment was echoed last year at the E-Magiciens conference held in Valenciennes, France, where Philippe Delvigne, director of the Industrial Design School (ISD), noted that the progress of 3D visualization and virtual reality was happening faster in Europe than in the US through the development of consortiums and centers for visualization, rather than big, proprietary systems. Delvigne points to the PerfVR effort, a collaboration to create standards for 3D visualization that would make the use of visualization centers easier.

Silicon Graphics (SGI), of course, is still in the game, and is also getting a lot of business in Europe (see “SGI Returns to Software,” pg. 32). Bob Pette, vice president of the visual business unit at SGI, says there tends to be far more small visualization centers throughout Europe than elsewhere. He notes that one of the largest efforts on this front was by CEA/DAM, a consortium spearheaded by the French Atomic Energy Commission. The group has built the CEA/DAM Ile de France Center in Bruyères le Châtel. This center is devoted to simulation and environmental surveillance, and is part of an effort to extend technology to other industries.

Even in Europe, that spirit of collaboration is shifting somewhat. “Security concerns,” Pette says, “are causing groups to build larger centers for their own use.” And, it’s worth adding that those uses are increasingly specific.

In France, Airbus has built an enormous CAVE to mimic the inside of the Airbus A380 main deck. The CAVE is 7.5 meters wide, 3 meters deep, and 3 meters high. It is built using 16 Christie projectors—four projectors each illuminate the ceiling, back wall, and floor; two projectors are used for each of the two smaller side panels. Projects are stacked to create a stereo pair for each channel.

The CAVE was built by Viscon, which also supplied the VisController II software to match up the images into a seamless, uniform image. RTT and VRcom supplied the software to produce the images. The CAVE gives visitors the feeling of being inside the A380 Airbus: It can demonstrate color schemes and change the seat arrangement for customers. The company is also interested in using the CAVE for engineering and ergonomics studies.

Multiuser Sites?
In the US, the visualization industry is being driven by the oil and gas industries. In oil and gas, there is little enthusiasm for sharing systems with outsiders. Nevertheless, Norwegian company Cyviz opened the Cyviz Technology Center (CTC) in Houston last May. The facility follows Cyviz’s original CTC, which opened in Norway the previous year. The company is also building a new CTC in Dubai, United Arab Emirates. Cyviz plans to host customers at the centers as well as use the locales as showcases for the company’s technology.

The center in Norway has three auditoriums and a conference room, and relies primarily on power-wall designs that range in resolution from 3.7 to 6.3 megapixels. Like Viscon, Cyviz supplies the software technology to create a seamless image from multiple computers, and the company works with different partners to build the system according to the customer’s requirements.

Projector companies Christie and Barco are especially active in this field, along with Hewlett-Packard, which provides the computers for many of these systems, including those at the Cyviz CTCs. The real trick in visualization is in getting the data from the systems onto screens in such a way that the imagery lines up and the colors are perfectly balanced so the illusion is not broken.

In theory, Headwave’s customers are the type of people who might use one of Cyviz’s centers, but Steve Briggs, vice president of systems integration, has his doubts. Headwave has developed compression technology that enables it to present enormous amounts of data to the user. The data gathered for oil and gas exploration is extremely dense—scientists might be looking through 7000 feet of water and 20,000 feet into the earth, as they were in the Gulf of Mexico when the enormous Jack Field discovery was made in 2006. As Briggs observes, “We’re staring at tens of terabytes of data.”

In this business, mistakes cost money—and lots of it. A deep-water oil rig costs more than $100 million. “Typically,” says Briggs, “companies will want to start a project, spending just a percentage of the entire project, and get to oil so they can start pumping to support the rest of it.”

Location is Key
Briggs believes visualization is critical but not always practical. He says that even when you have a visualization center on-site, it might not be that practical to use. Visualization centers have to be scheduled, and the data has to be prepared. In reality, he says, “people won’t even walk down the damn hall.”
Increasingly, teams collaborating across thousands of miles need to truly see an object. For instance, they need to see how a material will really look in a finished product.

Briggs continues: “Visualization centers are primarily for blowing the socks off executives or for those situations where an entire team wants to look at something.“

Comparatively, Headwave is seeing a lot more interest in deskside systems. Headwave’s software is based on a massively parallel architecture that enables it to take advantage of GPU clusters. This is one of the first products to take advantage of GP-GPU, or GPU-compute, technology that exploits the hundreds of cores in a GPU (for instance, the ATI R770 processor with its 800 processor cores, or the Nvidia GTX 200 with its 240 cores) to go to work on a much wider variety of number-crunching tasks. In fact, Headwave was one of the first companies to put Tesla, Nvidia’s heavy-duty, general-purpose GPU computing solution, to use (see “Digging Deep,” September 2008).

As Briggs notes, in spite of all the improvements in graphics hardware, one does not just fire up a 3D visualization and start collaborating away. Regardless of the display medium, the content has to be ported to the proper format that is conditioned for display. The work Cruz-Neira is doing at the LITE (Louisiana Immersive Technologies Enterprise) center is somewhat similar to the European model, though most often it is university personnel who are contracting with outside clients to create visualizations.

Understandably, Cruz-Neira takes issue with arguments that Europe is ahead of the US on the visualization front. She is an impassioned advocate for high-end visualization and a one-woman Johnny Appleseed for the technology. An example of her seed-planting is the $27 million, 70,000-square-foot LITE facility at the University of Louisiana in Lafayette (UL Lafayette), where Cruz-Neira is the executive director and chief scientist. It was built through the cooperation between the state of Louisiana, Lafayette Economic Development Authority (LEDA), and the university, and at least one of the major areas of research is oil and gas, in addition to medicine, weather prediction, entertainment, and industrial design.

The university has long been a regional champion, bringing high-tech practices to an area that has seen its local businesses face several crippling challenges over the decades, including the transformation from an agricultural district to one dependent on oil, only to be devastated by the departure of the oil industry from the US, followed by the long-lasting horrors brought by Hurricane Katrina.

The LITE center has three visualization venues that were built with SGI and Christie equipment—a six-sided CAVE with 1400x1050-resolution walls, a 174-seat Reality Center with a 37-foot screen, and a 20-room conference center. The venues are used by the LITE center’s partners and clients, as well as by the university, and for demonstration purposes to show local industry what can be done.

The systems are powered by SGI Prism systems, each one having 16 Intel Itanium 2 processors and six graphics pipes. LITE also has an immersive collaboration teleconference room that enables users to work across a high-speed fiber link. Last, but far from least, the center has an SGI Altix 350 supercomputer with 352 processors and 8tb of SGI InfiniteStorage SAN humming away in a back room to handle the heavy lifting.

It’s not just the hardware, though, that makes this possible. Cruz-Neira is able to accomplish what she has because the university is able to bring its resources to bear on particular problems. Most recently, UL Lafayette announced breakthrough work visualizing 3D data from mammograms to better understand breast cancers, and the center has also done groundbreaking work in weather studies and prediction. Furthermore, the LITE center is a tourist attraction of sorts, hosting regular open houses and conferences.
(Left) Showcase lets users put digital models in realistic environments complete with reflection maps for realistic reflections. The software scales from desktop to clusters. (Right) Customers can go inside the Airbus CAVE to get an accurate picture of the interior of their new plane before they take delivery.

Cruz-Neira believes her real work is to help companies build their own visualization centers because every company’s goals and problems are unique. “Yes, we do a lot of work with companies that use our centers, but we do a lot of work helping companies build their own systems,” she says.

Getting a Closer Look
Jeff Brum, vice president of marketing at Mechdyne, contends that the usefulness of a visualization center is closely related to its convenience. He believes that a system of centralized visualization centers is not as practical in the US because the country is so vast, and as a result, traveling to an outside visualization center doesn’t make as much sense as it might in Europe, where railways link many cities. Also, like Cruz-Neira, he finds that every customer presents a new challenge.

Mechdyne has licensed the CAVE trademark from the University of Illinois, and its Fakespace division worked on one of the largest CAVE projects in the US: the C6 upgrade at Iowa State (another project in which Cruz-Neira played an active part), completed in 2007. Mechdyne and Fakespace upgraded the university’s older system based on CRT projection that had a total resolution of 6.3 million pixels. The upgrade put 24 Sony 4k digital projectors pumping out resolutions of 4096x2160 to work. A cluster of 49 HP xw9300 workstations with dual processors and dual GPUs are driving the graphics. It’s the first system of its kind to support wireless tracking, and it is capable of presenting 100 million pixels. That little “redo” is reported to have cost Iowa State $4 million.

The C6 is being put to work on a variety of jobs for groups within the university and also for outside contractors. For instance, the US Navy is using the system for simulations. It renders specific topography and battle scenes, and enables control of virtual vehicles. In another application, researchers explore the makeup of plants on a cellular level. “You can fly into a genome and get a whole new appreciation of the biology of organisms,” says Brum, who has lost none of his enthusiasm for this technology.

In another project, Mechdyne has built a specialized power-wall application for paper-­product company Kimberly-Clark that replicates store shelves. It has a wide front wall and side walls for creating a virtual environment. According to Brum, Kimberly-Clark used to build actual shelves, exactly re-creating them from a retail chain and stocking them with actual products (or prototype packages) and competing products. Clearly, this was not cost-effective.
Headwave’s compression technology lets users, such as geologists, view enormous datasets, including well information.

As Mechdyne began installing high-definition screens in sites and working with industrial design customers, company officials were asked if the system could faithfully represent colors and even the stitching in the material. “We did a system for Ikea that uses multiple Sony 4k projectors. Their goal was to see the stitching in the seats of their furniture,” says Brum. But going to 4k means that the application has to be capable of displaying that resolution.

“You don’t want to just put in HD resolution and scale it up to 4k; the application has to be cluster-compatible,” Brum adds. And that’s something that providers like Mechdyne bring to the party. In fact, that ability to be cluster-compatible and to natively show resolutions up to 4k is going to become increasingly important on the high end.

The ability to visualize materials with perfect fidelity is an increasingly important area in industrial design, and it spans car design, furniture design, shoe design, and interior design. The ability to correctly reproduce materials and fabrics is already becoming the dividing line for some companies bidding for projects. RTT maintains that it first started working with car companies on this request, but is hearing from Adidas, the company’s client in the shoe business, as well as from furniture designers and airplane designers.

“It used to be enough to project textures. That’s no longer good enough,” says Snepvangers. “It has to be shaders, real-time shaders.”

If you think you’ve heard the magic word, you’re right. “Shaders” means graphics processors, and RTT has been working closely with Nvidia, using the company’s CgFX tools to create realistic procedural shaders that can be changed according to characteristics, such as color or thread count, to see what might work better. In the same way, RTT is also working on ways to reproduce paint.

“The paint companies are very interested in spectral rendering, trying to figure out when paint jars with an environment. It’s a hot topic,” says Snepvangers. “3D is going way beyond just 3D shapes; it’s now going into more of ‘what can I know through my eyes?’ The brain can imagine how material feels, what the characteristics of the material are, how it stretches and moves, how soft. We believe you can do a lot with your eyes.”

Autodesk is pursuing a strategy similar to RTT’s, but is building it from a different direction than RTT is. Autodesk acquired Alias, and along with it, the company’s industrial design software, Studio. The company added on to its portfolio with the raytracing company Opticore from Gothenburg, Sweden, and is putting its tools together in Showcase to create easy ways for companies to visualize designs. The offering is compatible with all CAD products and allows users to grab a model and put it, rendered, into an environment.

On the one hand, the firm is building easy desktop visualization. “Our goal is to democratize visualization, to make it very easy for the designer to get to a presentation quickly,” says Thomas Heermann, product line manager at Autodesk’s Manufacturing Solutions Division. On the other hand, however, Autodesk ran into the same opportunities that Mechdyne and RTT have seen on the high end of visualization—the ability to reproduce materials and textures.

The firm saw that on the marketing side as companies in the car business wanted to reproduce colors and materials; this has escalated to a point where companies are forming consortiums to try and figure out ways to measure color and material, and reproduce it with shaders. It’s an effort that is similar to the development of ICC standards, which have been developed to communicate color information between PCs and printers. “It’s the Holy Grail, to be as real as you can in the unreal world,” says Heermann.
Users gain a close-up view of complex data from the inside the C6 CAVE at Iowa State. The setup, which supports wireless tracking, is capable of presenting 100 million pixels.

Distributing the Vision
For the future, there are probably two models that will be used most often by businesses, and that will be to have some form of visualization on site—the best that can be afforded by the company—and also some networked collaboration. Brum, making the case for on-site systems, points to a study performed by Volkswagen several years ago in which engineers were asked how likely they were to use a visualization center if it was off-site, in another building, down the hall, or at the desktop. Like the people Briggs is working with at Headwave, most users said they might walk down the hall but they’d prefer to sit at their own desk.

The next step for visualization is to get all the realism of high-end visualization centers close to home. It seems clear now that for visualization to be most useful, it has to be down the hall and at the desk, but what hall, and what desk? And as Snepvangers points out, collaboration, increasingly, is not real time.

Globalization means that people are working at different times and in different places. RTT has developed its Picture Book software that lets users share images and access a model in RTT’s DeltaGen viewer, the firm’s raytracing application that takes advantage of the GPU for fast rendering. The Picture Book applications let users work with a server-based database to upload and access files at any time. This model also makes it easier for companies to work with outside suppliers.

In addition to RTT, Autodesk and Mental Images (now owned by Nvidia) have similar strategies. One of the most venerable companies in rendering in the entertainment sector and for industry, Mental Images has gone to work enabling long-distance collaboration based on its high-speed rendering technology with its Reality­Server, which maintains and processes content on the server. Images are streamed on-demand, giving access to huge datasets anywhere they are located. And, by never leaving the server, the system satisfies the need for security.

A lot of this work will be done at the desktop. Designers will put together quick demos of a proposed product with lighting, reflections, and beautiful surfaces, and they’ll share them for approval. They can be distributed on the Web or hosted in real time. There are numerous products popping up to do just this. BunkSpeed burst on the scene last year with fast raytracing to create beautiful raytraced images complete with high-dynamic range environments and reflections. At SIGGRAPH 2008, the company was fighting off a raft of competitors, including ArtVPS, StudioGPU, and Eon Reality.
Mental Images’ RealityServer enables users to upload models, information, drawings, and photos associated with a project; remote users can access that information at any time. The software scales to accommodate high-resolution raytraced images.

One competitive issue will be scalability. Like Mechdyne’s Brum, Autodesk’s Heermann says applications are going to have to accommodate increasing resolutions and clusters. RTT, Mental Images, and Autodesk are also taking advantage of the GPU. And wherever people are looking at visualizations and trying to make decisions, they’re going to want more.

The goal is to be able to put the same quality image on the desktop that one currently goes down the hall, across the city, or to Louisiana, or maybe over to France, to see.

Kathleen Maher is a contributing editor to CGW, a senior analyst at Jon Peddie Research, a Tiburon, CA-based consultancy specializing in graphics and multimedia, and editor in chief of JPR’s “TechWatch.” She can be reached at Kathleen@jonpeddie.com.

SGC Returns to Software

Building a visualization center is only one step on the way to accessing information and understanding it. The challenge is to get visual and related information into a format that is useful, and it may well be that just one pretty picture doesn’t tell the story.

Bob Pette, vice president of SGI’s visualization group, helps customers build visualization centers, and he’s been doing it for 20 years. During that time, the issues have changed. In the early days of advanced visualization, just getting 3D graphics to the display was the main challenge, and SGI developed key technology to enable that end, including customized pipelines and the first graphics processing units (GPUs). Today, says Pette, “graphics are no longer an issue.” Instead, people are overwhelmed by huge amounts of information.

SGI is trying to offer customers a better "view" of their large image sets by providing the data is a format that is useful and easy to understand through its new VUE technology.

For instance, a design project includes much more information than a CAD model. It may also include engineering and analysis information, technical drawings, and supply-chain information; or a site considered for oil and gas exploration will have geospatial data, satellite data, pipeline modeling information, and environmental data. In both cases, the information only makes sense if it is presented in a meaningful way.

For some time, the stumbling block for high-end visualization centers has not been technical, it’s been contextual. Visualization centers are not being increasingly used because it takes time to prepare data for presentation, and oftentimes customized software has to be developed for the application. Ironically, the whole point of visualization is to be able to make a decision quickly. The most valuable information is the information you can get to and use to make a decision as soon as possible.

SGI contends it has been working on this problem for several years and from several angles. The company builds Reality Centers for clients and supplies components for other systems. At the end of October 2008, the company introduced its VUE (Visual User Experience) software, which makes it easier to combine information from a variety of sources and deliver it to anyone on any platform. With the introduction of VUE, SGI is returning to software development, one of its core competencies, moved to back burners in the dark years of the company’s struggle to survive during bankruptcy and to start growing once again. Through the upheavals at SGI, CTO Eng Lim Goh has remained at the center of SGI, and he has continued to work on the vision of making visual information accessible across platforms and across the world.

Although SGI expects VUE to help sell Reality Centers and visualization hardware, SGI is selling VUE as a separate set of products for any visualization application. “These technologies,” says Pette, “enable OpenGL to run on any platform, including those of our competitors.”

There are other options, including proprietary systems built to order. Hewlett-Packard offers Remote Presentation Graphics (RPG) to customers that lets them access remote data on HP machines. Mental Images’ Reality Server is similar to VUE in that it’s agnostic about the platform.

Product Lineup
The VUE components include:

FusionVUE—Integrates information from various sources and platforms so it can be presented in VUEspace, SGI’s presentation environment.

SoftVUE—Renders visual information in software and does not require specialized graphics hardware.

PowerVUE—Renders information for large-scale applications in real time, taking advantage of CPUs and GPUs, depending on the platform.

RemoteVUE—Delivers visual information via Internet Protocol to remote locations on any device. Because the information is managed at the data center, RemoteVUE protects sensitive data.

EventVUE—Integrates diverse data sources and organizes it according to context or geospatial relationships. It can analyze real-time events based on established parameters, and issue alerts or highlight potential threats.

Although, SGI is now announcing VUE as a set of packaged modules, the company has been developing the technology for several years and making it available to customers. At this point, SGI has several customers already working with VUE components in the field. Halliburton is using a VUE-based system to monitor real-time drilling. NASA is modeling the effects of a dirty bomb blast in urban areas to help plan evacuation. And, in the really crazy category, the McLaren race team is able to run information from a race, pinpoint areas on the car that could be better streamlined, and rebuild a car in time for the next race—overnight if necessary.

SGI's VUE software modules are platform agnostic; therefore, not only do they run on the company's Reality Centers but also on systems from competing companies.

But perhaps the real power that new visualization technologies like VUE bring becomes apparent as they get broader rollouts into games, entertainment, and, of course, shopping.

Pette points out that the work the military is doing to train soldiers isn’t that different from a computer game. In both cases, VUE can offer more access to information and richer visual detail. VUE tools could be used to build better GIS databases that can be delivered to our car while we’re hunting for parking spaces. And, shopping for a car could get more interesting if one could pull up models on an iPhone and compare models and specs.

Listing the possible applications reveals nothing new. We’ve been talking about these capabilities for a long time. But, we haven’t had them yet. For SGI, the key is software, specifically open software. It’s a vision the company has held constant over the years, and it comes to life in VUE. –Kathleen Maher