MARKME: An AI-Powered Face Recognition and Bit-Level Data Modeling Framework for Scalable Attendance Management
Attendance management in educational institutions requires reliable identity verification, efficient processing, and compact storage of attendance records. This paper presents MARKME, an AI-powered attendance and identity verification framework that combines ArcFace-based face embeddings, FAISS similarity search, bit-level attendance representation, Redis state management, RabbitMQ asynchronous processing, and a FastAPI backend with a Flutter mobile interface. The implementation was evaluated using condition-wise face-recognition tests, API response measurements, concurrent processing tests, and storage comparisons. The reported recognition accuracy was 99% for good lighting and front-facing images, 88% in low light, 90% for a 45° side profile, 97% for a 90° side profile, 70% for masked faces, 85% with sunglasses, and 96% for different facial expressions. Storage evaluation reported 55 MB for traditional row-based attendance for 1,000 students per year compared with 240 KB using bitmask storage. These results indicate that the proposed architecture can reduce attendance-storage requirements while supporting automated recognition and asynchronous processing. The paper also discusses privacy, lighting sensitivity, network dependency, and deployment considerations.
Authors
- Pragati V. Thawani (ORCID: https://orcid.org/0009-0004-8984-4708)
- Nashrah Gowalker -
- Rakhi O. Gupta -
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24413539
- Primary Topic
- Face recognition and analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00