Smart Contactless Attendance System using Computer Vision and Edge AI

Project Abstract

This work presents a smart contactless attendance system that leverages computer vision and edge AI to provide reliable, privacy-conscious, and scalable tracking of participant presence in institutional settings. The system integrates real-time facial recognition, liveness detection, and privacy-preserving data handling to minimize contact points and reduce administrative overhead. A heterogeneous edge-computing architecture distributes processing across on-device inference units and local edge servers to ensure low-latency responses, robust operation in bandwidth-constrained environments, and resilience against centralized single points of failure. The core contributions include (1) an optimized face recognition pipeline tailored for resource-constrained devices, featuring lightweight neural network architectures, quantized models, and region-of-interest processing to accelerate inference while preserving accuracy; (2) a liveness verification module using multi-modal cues (blink, micro-movement, depth cues where available) to mitigate spoofing attempts; (3) a privacy-first data management framework that minimizes data retention, employs on-device feature extraction, and uses secure, auditable hashing for attendance records to satisfy regulatory and institutional standards; (4) an edge-accelerated attendance ledger with immutable logging and role-based access control to support auditing, reporting, and integration with existing student information systems; (5) an adaptive enrollment and recognition strategy that handles class rosters, opt-in consent, dynamic group assignments, and recurring re-enrollment to maintain high recognition rates across diverse demographics and changing environments. The methodology encompasses dataset acquisition and preprocessing, model training with transfer learning on domain-specific data, and deployment on heterogeneous hardware including micro-edge devices and local servers. Evaluation metrics focus on recognition accuracy, false acceptance and rejection rates, latency, throughput, energy consumption, and privacy compliance indicators under varying environmental conditions (lighting, occlusion, crowd density). A comprehensive comparative study against baseline attendance approaches demonstrates improved efficiency, reduced physical contact, and enhanced user experience without compromising security or privacy. The system includes an interactive administrator dashboard for real-time monitoring, attendance anomaly detection (e.g., duplicate entries, unattended sessions), and configurable alerting. To ensure scalability, the architecture supports multi-campus deployments, modular plug-ins for calendar integration, notification services, and analytics dashboards. Ethics and bias mitigation strategies are discussed, including diverse dataset curation, continuous model auditing, and transparent consent workflows. The results indicate high accuracy in face recognition with robust liveness verification across controlled and semi-controlled environments, while edge processing achieves low-latency performance suitable for large classes and high-traffic periods. The study also explores deployment considerations such as hardware heterogeneity, network topology, privacy-preserving key management, and maintenance protocols. Overall, the research demonstrates a viable, secure, and scalable solution for contactless attendance that can be integrated into existing academic and corporate ecosystems to improve operational efficiency, user trust, and data governance.

Project Overview

What This Project Is About

---

A practical project that explores how cameras and edge devices can automatically record attendance without touching anything. It uses computer vision to recognize people and edge AI to process data on local hardware, keeping information fast and private.

---

---

The Problem It Addresses

---

Traditional attendance methods require physical interaction or centralized servers, which can be slow, easily fooled, or raise privacy concerns. This project aims to provide a quick, contactless, and secure way to verify who is present in a group setting.

---

---

Objectives of the Project

---

1. Understand how cameras can identify faces or unique identifiers in a group setting.
2. Implement an edge AI system that runs locally on a device without sending data to the cloud.
3. Evaluate accuracy, speed, and privacy implications of the system.
4. Develop a user-friendly interface for administrators to view attendance records.
5. Ensure basic security measures to protect stored attendance data.

---

---

What You Will Do Step by Step

---

1. Study related work and choose suitable computer vision and edge AI approaches.
2. Set up hardware (camera, local computer or edge device) and required software.
3. Collect or obtain a labeled dataset for attendance scenarios (with consent and privacy in mind).
4. Train or configure a lightweight recognition model to run on the edge device.
5. Implement real-time video processing to detect and record present individuals.
6. Build a local database to store attendance events and generate reports.
7. Test accuracy, latency, and privacy protections; iterate on improvements.
8. Prepare a simple user interface and documentation for end users.

---

---

Expected Outcome

A functional, contactless attendance system that runs on local hardware, delivers fast results, and provides clear attendance logs while prioritizing user privacy.