Walter Sena

VISTA: Visual Inference for Spatio-Temporal Attention

An open-source community initiative for generating naturalistic, annotated video datasets to advance gaze tracking, cognitive assessment, and behavioral neuroscience.

🧠 Overview

VISTA is a framework and dataset initiative aimed at fostering more ecologically valid research in attention and cognition. We provide:

• AI-generated naturalistic videos simulating everyday scenes
• Human-curated scripts and interactions ensuring realism and task relevance
• Rich metadata annotations conforming to HED and FHIR standards
• Tools for researchers to create, customize, and annotate stimuli for cognitive and gaze-tracking experiments

This project supports explainable AI models and personalized neurocognitive assessment pipelines.

🔍 Key Features

• Naturalistic Video Stimuli: Real-world inspired visual scenarios designed for ecological validity
• Spatio-Temporal Attention Labels: AOIs, moving targets, and time-locked cognitive events
• Scripted & Versioned Content: YAML-based scene scripting
• HED + FHIR Metadata: Research and clinical interoperability
• Custom Stimuli Toolkit: (coming soon) Interface and CLI to generate and annotate scenes

🧪 Applications

• Benchmarking gaze tracking algorithms
• Modeling attention and cognitive load
• Remote neuropsychological assessment
• Autism and neurodevelopmental screening
• Aging and cognitive resilience tracking
• Multiple Sclerosis and brain fog

📜 Citation

VISTA: Visual Inference for Spatio-Temporal Attention (2025).
An open-source dataset for naturalistic gaze tracking and cognitive testing.
ubc.neurocognition.ai