Smart AI Attendance System Using ML Models
Independent project (Python, OpenCV): built the full pipeline from data collection through model evaluation, then deployed it as a full-stack web app with live tracking and an admin dashboard.
Technologies
Engineering Focus
Engineered through the complete development lifecycle: requirements analysis, database normalization (3NF), modular service architecture, concurrency control, and systematic testing.
01. The Real-World Problem Space
Traditional manual roll-calls and biometric touchpoints cause physical congestion, allow proxy attendance, and introduce error-prone paperwork.
02. Strategic Engineering Approach
Constructed a high-throughput computer vision pipeline using OpenCV and machine learning classifiers, packaging the complete solution with live tracking into a full-stack web app.
System Architecture & Service Flow
Video Stream Ingestion → Face Detection & Normalization → Feature Extraction → ML Classification → Live Dashboard
Data Pipeline & Preprocessing
Built the full pipeline from raw image capture and face alignment to data augmentation under varying light conditions.
Model Training & Evaluation
Trained and evaluated ML classifiers to achieve high precision and low false-positive rates on facial feature embeddings.
Live Tracking Engine
Processes real-time video frames with OpenCV to continuously detect and identify registered subjects.
Full-Stack Admin Dashboard
Provides real-time attendance logs, automated notifications, and historical export capabilities.
Technical Challenges Overcome
- •Optimizing frame processing throughput to sustain smooth 30 FPS video analysis without CPU bottlenecks.
- •Ensuring model robustness against motion blur, occlusion, and varying room illumination.
Demonstrated Outcome & Verification
Authored a complete research paper covering the system end to end, currently under review before journal submission.
Future Roadmap
Packaging the inference engine as an edge container running on lightweight embedded cameras.