Overview

AttentiveAI was a weekend project at the 2019 AT&T 5G Hackathon in Los Angeles. Our team built a mobile app that uses the phone camera to estimate how attentive each student in a classroom is, then overlays that score on each face in real time. The goal was to give teachers a quiet, in-the-moment view of the room, without pulling their attention away from students.

We didn’t win the overall event, but we made it past the first round of judging and took the prize for Best Use of IBM Technology.

What we built

• A cross-platform app in React Native and Expo (Android first) that detects faces and facial features (gaze, face posture, eye direction, eye openness) and turns them into an attention score
• An AR overlay showing each student’s score, plus space for context a teacher might want (sleep, hunger, stress)
• My part: a custom IBM Watson Visual Recognition classifier, trained on our own photos, to detect and count raised hands, with a backend on IBM Cloud
• Before building, we talked with teachers about their classroom challenges and shaped the features around what they told us

What I learned

• After the event, testers pushed back hard on how “creepy” it felt to track students this closely. The one idea that got real interest was turning the camera around to help teachers see and improve their own teaching.
• Work commitments made it hard to keep going, so we stopped there. I wrote up the full story, including a look at how attention can be measured, in Tracking Attention in the Eyes.