3D GAZE ESTIMATION IN INCONSISTENT CONDITIONS FROM LOW-RESOLUTION IMAGES FOR DESKTOP AND MOBILE DEVICES

Gaze

tracking

technology

provides information about the spatial coordinates of a person's point of attention in real-time using the distance to the

camera

, the position of the head, and the pupil of the

eye

. The

first

attempts at

gaze

tracking began at the end of the 19th century by a professor of psychology and education at Western Pennsylvania University to develop an effective

method

of the reading process when learning a language. () Nowadays,

gaze

tracking

technology

has found application in many fields, including medicine, psychology, marketing, and simulators.().

Although

ready-made solutions already exist, they have some disadvantages: the need for special equipment (Tobii

Eye

Tracker), high cost (Tobii Aware), or dependence on external

conditions

.

This

research is aimed at creating a

gaze

tracking

technology

accessible to everyone who has devices with a reasonable

camera

.

In other words

,

this

technology

will find application among the audience using computers, laptops, and mobile

phones

. According to Drewes H. (),

gaze

tracking already helps motor impaired people interact with a computer, and for ordinary people, it will make it more comfortable,

for example

, when typing text or as an alternative to a mouse or touchpad for mobile

phones

. Commercially successful areas,

such

as the video game market, are

also

interested in the emergence of a mass

Eye

tracking

algorithm

. Veronica S. claims that "Recent innovations in the video game industry include alternative input modalities to provide an enhanced, more immersive user experience.

In particular

,

eye

gaze

control has recently been explored as an input modality in video games." ().

In addition

to the wide application, computers and mobile

phones

have several other advantages: high computing power, high-resolution cameras, and their static position on the device, which allows for high-precision calibration ().

However

, the development of an accurate

Eye

tracking

algorithm

requires solving two critical problems: unstable

conditions

and poor image quality from the front

camera

. Due to the lack of infrared optical sensors in ordinary cameras, which provides robustness inexpensive devices (Tobii

Eye

Tracker 5) to changes in shooting environment

conditions

, the accuracy of

gaze

estimation will be affected by lighting,

camera

defocusing, as well as through view and scale changes. (). According to Shreya G., appearance-based and model-based methods are used in the development of

gaze

-tracking algorithms. (). Appearance-based models

use

a large amount of data and depend on the

method

of collecting

this

data (laboratory or in the wild),

therefore

they are extremely sensitive to changes in light and noise in the image ().

On the other hand

, model-based algorithms, according to Anuradha K., "

use

a geometric model of the human

eye

to estimate the center of the cornea, optical and visual axes of the

eye

and estimate the

gaze

coordinates as points of intersection where the visual axes meet the scene"().

Therefore

, the model-based

method

will be used to solve the problem with inconsistent

conditions

.

However

, the model-based

method

is extremely dependent on the

camera

resolution, which does not allow for the high accuracy of the

algorithm

. According to Fukuda T. () to reduce the cost of budget devices

such

as laptops and

phones

, manufacturers

use

cheap

camera

modules whose resolution is too low to

use

the

technology

with free head movements and independent face distance from the

camera

. To solve

this

problem, the

algorithm

will

use

the SRGAN model ().