Professional Project Brief – Revised for School Entrance Deployment Subject: Python Developer Needed: Automated Attendance System at School Entrance Using Facial Recognition (YuNet + InsightFace, CPU-Only, 900+ Students) I’m deploying a fully automated attendance system at the main entrance of a school (in the courtyard, near the gate). The goal is to log every student’s arrival time as they enter the campus — without any manual input — and automatically classify them as: • Present (on time) • Late (after official start time) • Absent (not detected by cutoff time) The system must run continuously during morning hours on a standard Windows 11 PC (CPU-only), using a single PoE camera installed at the entrance. ________________________________________ Non-Negotiable Technical Requirements 1. Open & Verifiable Models Only • Face detection: Must use the official OpenCV YuNet model (face_detection_yunet_2023mar.onnx from OpenCV Zoo). • Face recognition: Must use InsightFace with ArcFace (e.g., buffalo_l). • No private, undocumented, or black-box models (e.g., face_recognizer_fast.onnx). 2. Precomputed Embeddings for Real-Time Scalability • Student identities stored as numeric embeddings (.npy files), not raw images. • Live recognition via instant cosine similarity against precomputed database. • Must handle 900+ students in real time on CPU-only hardware (no GPU). 3. Accuracy in Real-World Entrance Conditions • Target: ≥95% true-positive rate under typical outdoor/semi-outdoor school entrance lighting (morning sun, partial shade, etc.). • Robust to common challenges: hats, partial occlusion, varying distances. • Configurable similarity threshold with clear inline documentation. 4. Transparent, Maintainable Code • Pure Python source code only —no .exe, setup.exe, closed SDKs, or obfuscated binaries. • Fully modular structure: • face_engine.py (detection + recognition) • attendance_logic.py (on-time/late/absent logic based on time windows) • database.py (PostgreSQL or MySQL) • camera_stream.py (PoE camera integration via OpenCV) • reporting.py (daily Excel + PDF exports) • Every function includes clear inline comments (purpose, inputs, outputs, parameters). 5. Automated Daily Workflow • System starts automatically at school opening time (e.g., 7:00 AM). • Continuously processes video stream, detects faces, and logs first valid detection per student per day. • At cutoff time (e.g., 7:30 AM), marks non-detected students as Absent. • Generates daily attendance report (Excel + PDF) by class/date. ________________________________________ Deliverables • Bill of materials: Recommended affordable PoE camera suitable for outdoor/semi-outdoor school entrance (with weather protection if needed). • Camera placement & network setup guide: Optimal height, angle, and lighting considerations for reliable detection. • Complete source code with requirements.txt, modular structure, and enrollment script (to register students from photos → embeddings). • Working demo video: Simulating real-world entrance scenario with multiple students walking in. • Step-by-step deployment guide for IT staff (Windows 11 setup, camera connection, service auto-start). • Database schema + sample queries + optional lightweight dashboard or REST API for attendance lookup. ________________________________________ Final Validation I will conduct a real-world test at the school entrance during morning arrival. The system must: • Detect and recognize students as they walk through the gate • Log accurate arrival timestamps • Correctly classify each as Present, Late, or Absent based on predefined time windows • Run stably for 3+ hours on a standard Windows 11 PC (CPU-only) Note: I will not accept partial code, undocumented models, manual capture steps, or closed binaries. ________________________________________ If you can deliver a transparent, open-source, production-ready system that meets these requirements, please reply with: 1. Your experience with InsightFace + YuNet in real-world deployments 2. Your approach to handle 900+ students under variable outdoor lighting 3. Estimated timeline & cost I’m looking for a developer who values verifiability, maintainability, and ethical AI — not just a quick demo.