An fMRI Dataset on Occluded Image Interpretation for Human Amodal Completion Research.

Area of Science:

• Cognitive Neuroscience
• Computer Vision
• Neuroimaging

Background:

• Partially occluded objects are common, yet challenging for AI.
• Amodal completion allows the brain to perceive whole objects from incomplete visual input.
• Existing research on amodal completion's neural basis is inconsistent due to small sample sizes and varied stimuli.

Purpose of the Study:

• To introduce the Occluded Image Interpretation Dataset (OIID), a novel fMRI dataset.
• To investigate human perception and brain activity during the interpretation of occluded images.
• To facilitate research into the neural mechanisms of amodal completion and individual differences.

Main Methods:

• Collected fMRI responses and behavioral data from 65 participants.
• Utilized a novel dataset (OIID) featuring images with varying degrees of occlusion.
• Employed functional Magnetic Resonance Imaging (fMRI) to capture brain activity.

Main Results:

• The OIID dataset provides rich data for analyzing brain regions involved in processing occluded visual information.
• The dataset allows for the examination of individual differences in functional responses to incomplete stimuli.
• This resource supports the identification of neural correlates of amodal completion.

Conclusions:

• The OIID dataset is a valuable resource for understanding how the brain handles incomplete visual information.
• Findings contribute to advancing theoretical knowledge of amodal completion.
• The dataset offers insights for practical applications in AI and visual perception research.