Visual Attention Saccadic Models Learn to Emulate Gaze Patterns From Childhood to Adulthood.

Area of Science:

• Computational neuroscience
• Developmental psychology
• Computer vision

Background:

• Traditional visual attention models provide static 2D saliency maps.
• Saccadic models predict eye movements (where and how observers look).
• Understanding age-related differences in visual exploration is crucial.

Purpose of the Study:

• To develop an age-dependent saccadic model that emulates visual exploration tendencies across different age groups.
• To investigate if saccade amplitude and orientation distributions serve as age-specific visual signatures.
• To compare the performance of the age-dependent saccadic model against state-of-the-art saliency models.

Main Methods:

• Collected fixation data from 101 observers across five age groups (2 y.o. to adults).
• Trained saccadic models using this data to capture age-specific viewing patterns.
• Analyzed the joint distribution of saccade amplitude and orientation for each age group.

Main Results:

• Identified unique saccade amplitude and orientation distributions as visual signatures for each age group.
• Generated age-dependent, human-like visual scan paths.
• Demonstrated that the age-dependent saccadic model significantly outperforms existing saliency models.

Conclusions:

• Saccadic models offer a flexible framework adaptable to emulate specific observer viewing tendencies.
• Age-specific saccadic models can accurately replicate developmental changes in visual attention.
• Computational modeling of visual attention can be tailored to specific demographic groups for enhanced understanding of gaze behavior.