Skip to content

Phenomena Reference

MindSight can detect and quantify a set of gaze-based social phenomena from video. Each phenomenon is an observable pattern of visual attention -- such as two people making eye contact or a group fixating on the same object -- that carries meaning in social cognition research.

Phenomena tracking runs on top of the core gaze-estimation and object-detection pipelines. You enable individual phenomena with their CLI flags, or activate all of them at once with --all-phenomena.

Phenomena Overview

Phenomenon Description CLI Flag Page
Joint Attention Multiple participants simultaneously fixating on the same object --joint-attention joint-attention.md
Mutual Gaze Two participants looking directly at each other (eye contact) --mutual-gaze mutual-gaze.md
Social Referencing A participant looks at another person's face then redirects gaze to an object --social-ref social-referencing.md
Gaze Following A participant shifts gaze in the same direction another person is looking --gaze-following gaze-following.md
Gaze Leadership Identifying which participant initiates shared attention episodes --gaze-leadership gaze-leadership.md
Gaze Aversion A participant actively avoids eye contact by looking away --gaze-aversion gaze-aversion.md
Attention Span Duration a participant sustains fixation on a single target --attention-span attention-span.md
Scanpath The sequential trajectory of a participant's gaze across objects --scanpath scanpath.md

Enabling Phenomena

Enable a single phenomenon:

python MindSight.py --source video.mp4 --joint-attention

Enable several at once:

python MindSight.py --source video.mp4 --mutual-gaze --social-ref

Enable every phenomenon:

python MindSight.py --source video.mp4 --all-phenomena

Each phenomenon may expose additional tuning parameters (window sizes, thresholds, quorum fractions). See the individual pages for details.