Reproduce “Smart Attendance System Usign CNN” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shailesh Arya, Hrithik Mesariya, Vishal Parekh
A webcam-based attendance prototype combines Haar-cascade face detection with a Siamese CNN that maps detected faces to 128-dimensional embeddings for identity matching. The system automates enrollment, recognition, MongoDB storage, and CSV export, reporting approximately 94% accuracy, although the evaluation lacks detailed dataset, error-rate, robustness, and security measurements.
Key insights
- The main contribution is an end-to-end application integration rather than a new recognition architecture: face detection, metric-based representation, classification, and record management are combined into one workflow.
- Embedding-based matching provides a compact representation for comparing live faces with enrolled users, but the system needs an explicit distance threshold or unknown-person rejection mechanism to avoid forced incorrect matches.
- Facial recognition introduces deployment requirements beyond recognition accuracy, including consent, biometric-data protection, liveness detection, and testing across lighting, pose, occlusion, and demographic conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.