// the take
A tkinter desktop app that takes attendance by recognizing faces with OpenCV's Haar cascade detector and an LBPH recognizer, writing one CSV per subject. A half-built Next.js/TypeScript frontend and Flask-style backend have since been bolted on alongside it. Fits students or hobbyists wanting a working classic face-recognition pipeline to study, not anyone needing a real attendance system.
The core pipeline (capture faces, train LBPH, recognize, log to CSV) is small enough to read start to finish in an afternoon, which makes it a decent teaching example of the classic OpenCV face pipeline. Haar cascade + LBPH needs no GPU and runs fully offline, so there's no cloud dependency or API cost. Responsibilities are split into separate scripts (takeImage, trainImage, automaticAttedance, show_attendance) rather than one monolith file.
The repo commits things a .gitignore should have caught: frontend/.next build cache, __pycache__ .pyc files, .idea project files, even a bash.exe.stackdump crash dump. The README tells you to manually edit hardcoded filesystem paths inside attendance.py and automaticAttendance.py before anything runs - there's no config file or CLI args. LBPH + Haar cascades are old and lighting/angle-sensitive, and there's no liveness check, so a printed photo or phone screen can mark someone present. The new Next.js/TypeScript frontend and Flask-ish backend look bolted on with no visible wiring to the original tkinter app, leaving two incomplete, seemingly disconnected systems in one repo.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →