Page 59 - 23_EETE_03

EE|Times EUROPE 59

Farewell to the ‘Paralyzed Cyclops’

FovoRender is, in effect, replacing any virtual
camera in a computer-generated scene with an
eyeball and an intelligent brain, he said. “By making
that one simple shift, you transform the scene.”
The startup’s initial market focus is 3D visualiza-
tion industries that make rendered images of cars,
interiors and buildings for brochures and billboards.
Game developers, virtual production, film produc-
tion and cinematographic special effects are also in
Fovotec’s sights.
Film director Karim Zouak, who has worked as a
visual effects artist for MTV, Channel 4, Nike, Sony,
Philips and Virgin, is already using FovoRender for
Unreal Engine in a CGI film project called
“Recognition.” “The ability to refine the depth of
Medieval vs. linear perspective, l.–r.: Duccio di Buoninsegna’s “Transfiguration” visual planes within wider shots really helps clarify
(ca. 1308–11); Pietro Perugino’s “Christ Handing the Keys to St. Peter” my visual storytelling and deepen emotional engage-
(ca. 1481–82) (Source: Wikimedia Commons) ment,” Zouak told an interviewer in a recent blog. 2
The next step for Fovotec is to target its core
A medieval religious picture, for example, might have a larger Jesus IP at GPU pipelines in order to process images in real time. This is
or Madonna in the center, with smaller saints or angels on the periph- particularly important for applications with a feedback loop between
ery. This is exactly how you and I see, Pepperell and Burleigh said. the viewer and machine so that the GPU can process images according
We have an exaggerated, central focus on what’s important, so those to head or eye position. Fovotec is already working within GPU giant
objects loom large in our field of vision (think of the huge moon). Nvidia’s Inception program for startups and is collaborating with
Around the edges, things seem smaller and less distinct. We perceive Intel to integrate FovoRender into OSPRay, an advanced path-
them using peripheral vision, which watches out for trouble in line tracing–based renderer. Adding FovoRender to GPUs will be “like
with our “fight or flight” instincts. adding Dolby for the visual space,” Burleigh said.
“Human vision is fixation-driven; linear perspective is fixation- Looking further ahead, there’s potential to make films with
agnostic,” Pepperell said. “Human vision is also dynamic, so every next-generation depth-sensing cameras that work more like human
time you focus on something new, it changes and gets bigger and eyes thanks to curved CCDs and adaptive lenses. “The human retina
more distinct. Our vision is constantly adapting in real time.” has depth, so you can see around things to a certain extent, even
To achieve this natural viewing experience on a computer screen, using one eye,” said Burleigh. “Cameras with depth sensing are
FovoRender generates and modifies parameters like spatial coordi- perfect for FovoRender, as the software works with full 3D informa-
nates, color, texture, surface-mapping data, motion data and object tion and can display that depth-based content on a 2D device to best
identification data of 2D images of 3D scenes. Head and eye move- effect.
ment can also be used to control the core algorithms. “It’s amazing how much depth you can get on flat images,” he
Integrated within a 3D content creation tool like Unity or Unreal added. “You can trick our visual system with natural depth cues to
Engine, FovoRender provides a series of preset projections based on the point where you can get close to the depth of VR with a normal
the principles of natural vision. A multitude of adjustment functions screen. How far can you go?”
are accessed by a slider interface and can be adjusted to preserve the It’s a question that many in the industry will be asking, too. ■
natural shape of objects and reduce distortion as a virtual camera
moves through a 3D scene. REFERENCES
Alternatively, 3D models can be shown in FovoRender and directly 1 Ruta, N., Burleigh, A., and Pepperell, R. (August 2022). “Space and Scale in
linked to sensors for detecting the motion and position of a human Medieval Painting Reflects Imagination and Perception,” Gestalt Theory, Vol. 44,
viewer’s eyes, head and body. As the human viewer moves about, the No. 1-2, pp. 61–78. bit.ly/3RibJt3
geometry of images is automatically modified, with objects getting 2 Fovotec. (Nov. 11, 2022). “Fovotec collaborates with top Film Director Karim
bigger, say, as the viewer moves closer to the screen. “It means we can Zouak.” Blog post. bit.ly/3XT4TMY
experience digital and remote spaces as though we are actually there—
often without VR or AR equipment,” Burleigh told EE Times Europe. Christine Evans-Pughe is a contributing writer for EE Times Europe.

Image with FovoRender (left) and with standard linear perspective (right) (Source: Fovotec)

www.eetimes.eu | MARCH 2023