FocusLock

Inspiration

Every student knows the struggle — you sit down to study, and within minutes, focus drifts. Every teacher knows the frustration — running fair online exams feels impossible.

We wanted to fix both. FocusLock was inspired by the idea that awareness > control. Instead of building another strict proctoring tool, we wanted an AI that understands attention — one that helps students build discipline and helps teachers uphold fairness, calmly and transparently.

What It Does

FocusLock uses your webcam to track subtle behavioral cues — eye movement, head direction, and phone presence — to calculate a real-time focus score.

For students: It’s a digital mirror that shows when your mind drifts, helping you build concentration over time. For teachers: It detects cheating cues and phone use during online exams, ensuring fairness for everyone.

It’s not surveillance — it’s awareness.

How We Built It

We built the backend with FastAPI, integrating:

MediaPipe FaceMesh for real-time facial landmarks and head-pose estimation
YOLOv5 for phone and object detection
WebSockets for live video frame streaming from the browser to the backend

Each frame is analyzed and scored in real time using a custom smoothing algorithm:

[ \text{FocusScore}_t = \alpha \cdot \text{FocusRaw}t + (1 - \alpha) \cdot \text{FocusScore}{t-1} ]

Challenges We Ran Into

Running heavy CV models like MediaPipe and YOLO in real time without crashing the browser.
Streaming video frames via WebSockets efficiently enough to handle real-time inference.
Maintaining smooth focus score transitions that feel natural, not robotic.

Balancing all this in a single web app — with minimal lag and high reliability — was a huge engineering challenge.

Accomplishments We’re Proud Of

Built a fully working prototype that runs computer vision models and live feedback in the browser.
Created a real-time focus scoring system that feels intuitive and stable.
Designed a dual-purpose system that helps both students and teachers — bridging personal growth and educational integrity.

What We Learned

We learned how to simplify complex AI systems for real-time web use. We learned that building focus isn’t about restriction — it’s about reflection. And that smooth user experience is just as hard — and important — as model accuracy.

What’s Next for FocusLock

Turning it into a browser extension and mobile SDK for schools and learners.
Adding LLM integration to give personalized focus insights (“You lose focus after 8 minutes — try shorter sprints”).
Expanding detection to posture, emotion, and engagement — creating a true AI study companion.