For Stealth HD’s Video Software, A Panorama of Applications

7/10/12Follow @wroush

In 2007, Bill Banta got discharged by the U.S. Navy’s fighter pilot training school because he failed an eye exam. Today, his Silicon Valley startup is building video software that could help pilots by giving them eyes in the back of their heads.

You could call that irony, or serendipity. After all, if Banta had stayed in the Navy, he wouldn’t have been able to apply for a job at Apple, where he ended up running the team that located camera parts for the iPhone. He wouldn’t have gone on to Stanford Business School, and he probably wouldn’t have developed his idea for Stealth HD, whose software stitches the signals from up to four cameras into a real-time, high-definition, fully panoramic video feed.

That’s a technology that could be useful not only to pilots, but to soldiers, athletes, security guards, sports broadcasters, mapping companies—basically, anyone who could do their job better if they had a 360-degree view of the world. Stealth HD’s system works best with four cameras arranged at right angles; it feeds video to a customized viewer on a PC screen, where a user can pan right or left in real time to examine any part of the view. (It works with recorded video too.)

“Imagine a military situation like an armored vehicle with big blind spots,” says Banta. “This would be like poking your head out the top and being able to see everything.”

This May, Stealth HD won the “Product Showcase” portion of the BASES 150k Challenge at Stanford, where Banta is in the middle of the two-year MBA program. Together with co-founder Ross Biestman and former Apple engineer Geoff Donaldson, Banta is busy fundraising for the startup, while at the same time recruiting beta users in markets such as sports, entertainment, and military surveillance. He says deals to deploy the prototype technology are in the works with ESPN, Red Bull, and a couple of NFL teams.

The idea of panoramic moving images isn’t new: If you ever visited the Circle-Vision theater at a Disney theme park, you’ve experienced a primitive version of the concept using nine separate film projectors. Companies such as Immersive Media and yellowBird make computer-age version of Circle-Vision movies, with the added bonus that a user can control the point of view in the finished videos using the mouse or touchpad, the same way they would with a VR panoramic photograph.

But making a 360-degree video is hugely processor-intensive, which means an Immersive Media or yellowBird video has been through a lot of post-production work before you ever see it on the Web. Stealth HD’s focus is on making that processing so fast and efficient that the graphical processing units (GPUs) inside a standard PC or laptop can stitch together a panoramic image in real time. “We are trying to develop a solution that is fully real-time and fully HD, no matter where you are looking,” says Banta. (See the video on page 2 for a demo.)

The idea for Stealth HD hit Banta and Biestman during a 2010 ski trip to Squaw Valley. But they’ve done a lot of slaloming since then.

“We were sitting in line at the lift and we counted 35 people wearing some type of camera on their helmet, like a GoPro or a Contour,” Banta says. “We decided they all looked ridiculous, but the idea was really cool, so we should come up with a way to embed the camera inside the helmet.”

In a further twist, Banta and Biestman decided one camera wasn’t enough. Most skiers are so bad at aiming their helmet-cams that “it’s not a pleasant viewing experience afterward,” Banta says. “We started toying with the idea of having multiple cameras, so you wouldn’t have to worry about which direction your head is pointed.”

Banta already had quite a bit of experience building stuff. As a mechanical-engineering graduate just out of college, he’d started a company to make environmentally friendly foam blanks for surfboard cores (as had others). After that washed out, and after he got ejected from flight school, he joined the camera group at Apple, at the exact moment when the company started putting tiny camera assemblies inside every iPhone and iPod touch. “We went from doing $65 million a year in parts sourcing to well over $1 billion, and I kind of got to take it through that process,” Banta says.

It wasn’t until after he’d left Apple, worked briefly for Square, and entered Stanford that he got a chance to pursue the pano-ski-helmet idea. But almost as soon as Banta and Biestman began speaking with potential investors, they realized the project wasn’t fundable. “Everybody we talked to said, ‘A, that’s not a big enough idea, and B, don’t underestimate how expensive it’s going to be to generate all this custom hardware and mass-produce it when GoPro and Contour already have a big chunk of the market.’”

By that time, late 2011, the startup had already begun developing the software they knew they’d need to stitch together video signals from multiple cameras. The solution was obvious: Make the software the product. “We realized that we should do was focus on building a platform that allows people to generate and view this type of video,” Banta says. “That was a difficult shift, personally, because my expertise is hardware and we were getting rid of the hardware piece. But it definitely made sense for the company.”

Banta says the special sauce in Stealth HD’s technology is spread across three different areas. First, there’s the parallel processing that needs to be done to analyze the video from four separate cameras, identify the overlap, and stitch the images together so that the seams aren’t noticeable, all at 20 frames per second (the startup hopes to hit a full 30 frames per second by the end of the summer). Second, the software has to do a lot of additional processing to ensure that color saturation and exposure levels are consistent all the way around a 360-degree image. Third, there’s the problem of wrapping the image around on itself, forming a true cylinder. After all, the underlying image file is just a big, flat rectangle that’s 1,080 pixels high and 7,680 pixels wide (4 x 1920). “For a long time, we were able to get three of the seams to look really good, but the fourth was always bad because it was the one with the outside edges,” says Banta. But the problem was eventually solved—see the demo video below.

The startup’s May victory in the Stanford entrepreneurship competition brought a $20,000 prize. “That was a big validation,” says Banta. “Now it’s a matter of moving forward and figuring out who we want to partner with, finding investors, building out the team, and building out the platform.” In other words, everything.

Having sidestepped the hardware business, Banta thinks Stealth HD will end up earning revenue as a cloud service provider. “Say you have a panoramic video stream that you want to create,” he says. “You stream it to our cloud, we process it and stream it back to you, and you pay us X dollars per month, and you do whatever you want with the video.”

That “whatever” could range from athletic development to surveillance. A quarterback with panoramic cameras in his helmet, for example, could review Stealth HD’s video after a game to figure out where the defensive lineman who sacked him came from. Even more intriguingly, football fans subscribing to some future enhanced broadcasting service might be able to call up the helmet-cam video on their tablets as a premium add-on.

On a more practical level, Stealth HD’s technology could simplify video surveillance. “If you have a power plant or a military base, instead of setting up 50 cameras to monitor a perimeter, why not just set up one at each corner?” asks Banta. It might even be sensible to mount omnidirectional cameras on drones like the Predator, to supplement the views available to remote drone pilots. “There is a lot of value to these super-high-resolution zoom cameras, but the analogy one Air Force guy gave us is that it’s like looking through a straw,” Banta says. “At the same time you need situational awareness, so you can analyze what is going on around the target.”

Banta says Stealth HD has no commercial partnerships to announce, but it sounds like there are quite a few in the formative stages. If the technology does take off, Banta would probably feel a lot better about missing his chance to fly Navy jets. “Wouldn’t it be ironic if this ended up making a much bigger contribution to the military than I could have as an individual pilot?” he says. “If it works out that we can save lives or help people make the right decisions in real time, then that is awesome.”

Wade Roush is a contributing editor at Xconomy. Follow @wroush

By posting a comment, you agree to our terms and conditions.