FocusTrack

Real-time focus monitoring system using computer vision and deep learning. Tracks attention, drowsiness, and blink rate during study/work sessions via webcam.

How It Works

1. Emotion Recognition — Custom 7-class CNN trained on FER2013 (28,709 images), 59.28% test accuracy (human-level baseline on FER2013 is ~65%)
2. Eye Tracking — Haar Cascade classifiers detect both eyes, monitor blink rate and gaze centering
3. Focus Classification — Rule-based engine maps emotion + eye state → Focused / Distracted / Drowsy
4. Session Dashboard — Auto-generated visualization with timeline, productivity score, and blink rate graph

Stack

Python TensorFlow OpenCV MediaPipe Pandas Matplotlib

Model

• Architecture: Custom CNN — 3 conv blocks, batch normalization, dropout, 512→256 dense layers
• Dataset: FER2013 (48×48 grayscale, 7 emotion classes)
• Test Accuracy: 59.28% | Test Loss: 1.058

Setup

git clone https://github.com/0AnshuAditya0/focus-track.git
cd focus-track
pip install -r requirements.txt

Usage

# Run live focus tracking session
python focus_tracker.py

# Generate dashboard from saved session CSV
python dashboard.py session_YYYYMMDD_HHMMSS.csv

Press q to end session and save data.

Sample Session

Metric	Value
Duration	1m 18s
Productivity Score	83%
Focused Time	83.8%
Distracted	13.5%
Drowsy	2.7%
Avg Blinks	0.0/min
Top Emotion	Neutral

Author

Anshu Aditya — GitHub