چكيده به لاتين
Human Gaze Estimation consists of Eye tracking and providing the computational model for gaze estimation. Human gaze estimation plays a crucial role in expressing a person’s desires, needs, cognitive processes, emotional states, and interpersonal relations. The importance of eye movements to the individual’s perception of and attention to the visual world is implicitly acknowledged, as it is the method through which we gather the information necessary to negotiate our way through and identify the properties of the visual world. Human gaze estimation has many applications in behavior and attention analysis, human-computer interaction, etc.
There are been numerous methods for tracking eyes and estimating gaze, but despite active research and significant progress in the last 20 years, gaze estimation remains challenging due to the individuality of eyes, occlusion, variability in scale and head pose rotation, location, and light conditions. In this research, we have investigated recent methods and presented a new architecture, based on convolutional neural networks and probabilistic graphical models. We have used EYEDIAP and MPIIGaze datasets and did multiple experiments. The results shows Mean Error degree of 7.5 and 6.4 for mentioned datasets, respectively.