Belarus-based startup Banuba only came into existence at the start of 2016, but it has already made a name for itself in the world of augmented reality. Now, backed by venture capital, the firm has companies like Instagram and Snapchat set in its sights.
Co-founder Yury Hushchyn mostly works on the research and development side of the startup, overseeing a team of about 30 researchers. Speaking with Techworld, he explained how he had previously worked with the other co-ounder, Viktor Prokopenya, on a high-frequency trading project, using analytics based on artificial intelligence and machine learning.
They then applied this accumulated knowledge to creating computer vision applications for mobile devices. "We believed we could create something that was unique and state of the art in many different aspects," says Hushchyn.
According to Hushchyn, Banuba's version of AR may be different to what we would generally associate with augmented reality, such as you might expect from immensely popular games like Pokemon Go. This vision, Hushchyn says, can be abbreviated as SLAM - Simultaneous Location and Mapping - AR, which relies on feature points to recognise objects and scenes.
"We take another approach, which is our definition of augmented reality: we are focusing on the front camera-based AR," says Hushchyn. "Which means that we augment selfies, with additional objects and we are making mobile applications that are context aware.
"We focus on two big areas, one of them is precise and high-performance face tracking and the other is the high performance and precise face segmentation."
The latter means that they can isolate features of the face including hair, brows, eyes and lips.
The simplest way to describe the group's current product is similar to that of Snapchat and Instagram facial filters. However, Hushchyn says that Banuba's product goes further. "It is not limited only for framing but instead it uses imitation as well so it's going actually way beyond a typical AR," he says. You can watch the advanced AR technology in use here.
Another selling point for the startup is in tailoring its technologies to extremely low-end and low-power devices. Hushchyn says that at an upcoming conference in Barcelona they'll be showcasing the technologies running on Android Go devices.
"We actually try very hard to bring hardware requirements as low as possible to make the technology accessible," he says.
In this low-power environment, where there is not access to sufficient high-speed internet to stream video, AR can be used to support avatar-based video communication instead.
Hushchyn says the group's nascent technologies have the potential to create an entirely new class of applications. Some of these are practical, such as Banuba features that let user try on fashion items like glasses, while others are focused on pure entertainment.
One of the more obscure applications addresses the optical distortion inherent in selfie-taking.
Due to the angle and proximity of the face to the device, taking a selfie can contort, in subtle ways, the resulting image - meaning that a photo taken with our front-facing cameras makes us look slightly unlike we do when simply looking into a mirror.
Hushchyn says that this can make people feel self-conscious when using video conferencing, and that AR can be used to correct this effect - as well as incorporating filters and layering different effects on top of the video images.
One form of this might be that certain effects are triggered by facial expressions - for example, a smile triggering a heart emoji. "So imaging is automatically inserted based on your facial expression, which helps to deliver your message better and your state of mind is clear in that communication," explains Hushchyn. He says this adding this extra context into communication is one of the startup's main aims.
Eye tracking is another important frontier for the group, with user of Banuba apps able to control the screen just by focusing their eyes rather than manually interacting with it.
In terms of AI, Hushchyn says that the company develops it for both in-house use and to be integrated into their consumer experiences too. An example of one of these is the Banuba Filters service which delivers customers 10 personalised filters everyday to dress their selfies in.
But why has AR been so slow to take off in any practical way? According to Hushchyn, the twin development of better algorithms and the greater computational power of devices today means that the area is set to take off in a big way.