How projectavsm/digital-facial-recognisation-attendance-system works
Project type + purpose (from README + tree)
This is an edge-deployed, full-stack IoT/computer-vision “appliance”:
• Runs on Raspberry Pi 5 with Pi Camera Module 3
• Performs real-time face detection + embeddings + classification (MediaPipe + FaceNet + scikit-learn RF)
• Provides a Flask web dashboard (templates/static)
• Persists attendance + directory data in MySQL (schema.sql)
• Exposes the dashboard securely via Cloudflare Zero-Trust Tunnel
• Auto-starts on boot via systemd with stability/boot ordering and self-healing
bridge.py obtains public URL and emails admin (SMTP)
5. start_system.py launches run_pi.py and/or Flask server (depending on your implementation split)
Student identified → DB transaction
• Check presence uniqueness constraint (composite key)
• Insert attendance record if not already present today
Camera pipeline is not OpenCV-first
Explicitly label: rpicam-vid YUV streaming → conversion → inference, because the README says it’s intentionally bypassing V4L2/Libcamera locks.
Dual source of truth for “student directory”
• MySQL student table(s)
• /dataset folder structure
And show the admin directory “smart portal” scanning both (README: dual-lookup mapping Student IDs to physical folders).
Remote access is not direct port-forwarding
The “public URL” is through Cloudflare tunnel, not NAT. Show Cloudflare as a first-class component, not a footnote.
Concrete component list to include (maximally decomposed, based on repo)
• attendance.service (systemd unit) (even if file isn’t in repo; README states it)
• start_system.py (boot orchestration)
• bridge.py (tunnel + URL + email)
• run_pi.py (camera loop + server integration)
• Camera capture: rpicam-vid YUV stream
• Frame conversion: YUV → BGR
• MediaPipe Face Detection
• FaceNet Embedding Generator
• RandomForest Classifier
• StudentID ↔ dataset folder resolver (dual-lookup)
• Flask routes
• SQLAlchemy session/ORM
• Auth/login (templates indicate login.html)
• camera_mark.js (mark attendance flow)
• camera_add_student.js (enrollment flow)
• dashboard.js / admin.js
• hardware.py (GPIO/buzzer/LCD)
• MySQL (schema.sql, unique constraints)
• Filesystem dataset (/dataset + symlink to static)
• Cloudflare tunnel
• SMTP/email provider
• Admin browser