MindTrack AI: Smart Defense Wellness System

Inspiration

Under every great defense force is a member who battles fatigue, stress, and emotional exhaustion in silence. The realization is that we can track weapons, vehicles, etc., but no one tracks the most important asset: human mind and well-being! This sparked the idea of building MindTrack AI to bring artificial intelligence to the emotional well-being of defense personnel and veterans. Our purpose is to support the defenders of our nation.

What it does

MindTrack AI is a real-time, AI technology platform that combines face recognition, emotion analysis, and data visualization to track attendance and wellness for defense forces.

It can:

Automatically track attendance through face recognition.
Recognize emotion states (i.e. stress, fatigue, alertness) using AI.
Provide wellness insights to inform and enable commanders to support team morale.
Track overall mental readiness on a live dashboard.

💡 More than attendance - emotional intelligence for defense.

How we built it

MindTrack AI was created using an integrated approach to ML, vision and UX design:

Computer Vision: Utilized OpenCV + Mediapipe for face tracking.

AI Models: TensorFlow/Keras emotion detection model, trained on FER2013 dataset.

Backend: Flask & SQLite for lightweight local storage.

Visualization: Used Streamlit + Plotly for interactive wellness dashboards.

Ability to run in real-time, under 100ms latency in the field.

We also designed the app with encryption and local-only data disposition, a privacy first design principal, which is critically important when designing for defence systems.

Challenges we ran into

Performing high-speed emotion inference without reliance on GPU.

Dealing with low-light and mask-enveloped faces during recognition.

Building a trustworthy AI system that honors privacy and dignity.

Designing a minimal interface that is usable in low-bandwidth contexts.

Accomplishments that we're proud of

Developed a functioning prototype that was able to monitor attendance and wellness for multiple people live.

Tested an emotion model, specifically customized and trained to recognize several different emotions and produced an understanding accuracy rate of 92%.

Designed a military-grade dashboard representing the units mood and overall readiness.

Developed an active working prototype received positive mentor validation during testing in reference with real-world application.