Face Track Pro:
AI Attendance System
A single-file face recognition attendance system that identifies up to 20 students simultaneously per frame — using CNN (GPU) or HOG (CPU) detection, with SQLite storage, Tkinter UI, and Excel/PDF/chart exports.
Face Recognition Attendance — Tkinter Desktop UI
Up to 20
Faces per frame
CNN + HOG
Detection model
CUDA / dlib
GPU support
Completed
Status
Why I Built This
Universities and companies in Kurdistan still use manual attendance — paper sheets, roll calls, and prone-to-error processes that waste class time and allow proxy attendance.
Existing solutions were either expensive commercial software or simple QR-code apps that required students to actively scan something. A truly passive, camera-based system was missing.
System Design
Face Detection: dlib CNN model with automatic CUDA GPU detection — falls back to HOG on CPU-only machines. Processes every 2nd frame at 50% scale for performance.
Storage: SQLite with 5 tables — students, teachers, classes, class_students, and attendance records with session tracking.
UI & Exports: Tkinter desktop GUI with Matplotlib charts, Excel (openpyxl) and PDF (fpdf) export, and unknown visitor photo logging.
What It Can Do
20 Faces Simultaneously
Processes up to 20 faces per camera frame — handles full classroom size in one shot without individual scanning.
GPU-Accelerated CNN
Uses dlib's CNN face detector with CUDA support for real-time speed. Automatically detects GPU availability and falls back to HOG on CPU.
Complete Analytics
Matplotlib attendance charts, Excel spreadsheets with color-coded formatting, PDF reports, and unknown visitor photo archive.
SQLite Schema
Explore Related Projects
The Gesture Recognition Engine uses the same AI/CV stack — real-time body tracking with a custom ML model.