Evaluation of Eye Focus using Haar Cascade Classifiers and Template Matching in Video

A new technique of extracting and evaluating the eye behavior property (eye focus) from an input video was proposed and presented in this video. The system works most accurately for frontal faces and good view of the eyes (no obstruction and well-lit conditions).

The solution proposed involves three main steps:
1) Face Region Localization;
2) Eye Regions Localization; and
3) Template Matching.

The system is designed to adapt with the input video to avoid using generic eye templates which is not optimized for our native complexion and appearance. The system displays the results in a graph containing the history of the movement of eye focus that can aid, for example, behavioral scientists in their procedures.

Legend:
Red box - Face Region
Blue box - Eye Regions
Green box - Eyes looking "in front"
Light blue box - Otherwise

Acknowledgements:
I would like to thank the volunteers on this video. However, I will not name them for privacy and security reasons.

Category:

Education

Tags:

• eye
• eye focus
• eye location
• haar
• haar cascades
• template matching
• opencv
• localization

[rockypaladinknight]

Wow !!! This is what I looking for. Now I currently work on project that can tell whether the user focus on something by using eyes focue. Could you please send me what your image use for Template or tell me where to find them ? Thanks for Advance :)