Magic Leap recently revealed a new machine vision technique for its long awaited standalone AR headset. Edgy Labs sorts through the rumors and looks at what’s on the horizon for mainstream Mixed Reality entertainment.
After about six years of existence, 4.5 billion dollars of market valuation and one eye-popping marketing campaign, what does Magic Leap really have in store?
6 Years and $1.4 B later, Still no Product in Sight?
Magic Leap initially wowed us with its mixed reality tech that would allow superimposing 3D elements on to the user’s environment in real-time.
Reminiscent of Microsoft’s marketing approach for the Hololens, Magic Leap generated quite a bit of buzz through several mind-blowing video releases last year.
But, after more than half a decade of waiting, the buzz is turning sour.
Between a number of voices in the industry that are cynical about the actual capabilities of the company’s AR technology and law suits over the company’s work atmosphere, the secretive AR startup has evidently experienced some turmoil as of late.
For example, tech news outlet The Information argues that AR technology being developed by the tech startup may not match up with what the demo videos presented so far might have us believe.
Amid mounting criticism directed at the actual development of Magic Leap tech, the site stated that the videos produced were not accurate representations of the state Magic Leap’s technology.
Magic Leap has faced accusations that the demo videos benefited largely from special effects for a final rendering that were very different from what testers had actually experienced.
The company also recently settled a sex discrimination lawsuit following accusations of sexism brought by Tannen Campbell, the former vice-president of strategic marketing.
Magic Leap is Taking its Time, for a Reason
But fear not!
Last week, the Florida-based company’s researchers published a paper (Toward Geometric Deep SLAM), in which they describe a novel machine vision technique that runs on dual deep convolutional neural networks (CNNs).
Specifically, the system is made up of twin CNNs known individually as MagicPoint and MagicWarp and is intended for the startup’s coming stand alone AR headset.
GreenLights Insights’ Senior Tech Analyst explains the new Magic Leap approach:
“MagicPoint processes 2D images to identify corners, and assigns points to each identified corner. Then MagicWarp compares the before and after point locations to determine and predict movement. Each convolutional neuron processes data only for its receptive field, allowing CNNs to interpret movement in the images. The result, according to Magic Leap is a system that’s “fast and lean, easily running 30+ FPS on a single CPU. The Magic Leap vision tech may help minimize the number of sensors needed in their headset, especially compared to an estimated 11 or more sensors in the Microsoft Hololens.”
Magic Leap is developing a novel machine vision technique for its standalone AR headset.Click To TweetMagic Leap Mixed Reality Glasses
While the ultra-secretive company still doesn’t seem ready to market its Mixed Reality (MR) glasses, it is, however, showing signs of advancement contrary to impatient criticism.
A September 2015 patent filing confirms that the start-up is still developing its glasses, which are (in true Magic Leap fashion) largely anticipated but without many details on progress or release.
But, we’re not flying completely blind, so to speak. We do know a little more about the appearance that Magic Leap’s mysterious MR device might ultimately take.
Spotted by Business Insider, the patent filing has just been approved and published after nearly two years and contains images showing a pair of glasses that are thinner than expected.
The glasses feature oval frames with a total of four cameras with two on either side.
If these new reports accurately represent the state of Magic Leap MR development and innovation, we may be on the verge of an AR-driven world previously only seen in Sci-Fi. And in Magic Leap promo videos, of course.
Could vision be manipulated and a mixed reality produced to other people? Cameras and even human eyes would be cheated with wrong signals.
A perfect hoax may be difficult or impossible. More important is whether people want to protect their privacy against face identification of many applications. Ability to send even some scrap to cameras might help.