International Journal of Scientific Research and Engineering Development

In India Colleges and institutions depends on manual roll number call or attendance sheets Wasting almost 10 to 15 Minutes of every lectures and also get fake attendance This problem Lead to the Development of the VisionMark a Artifiial intelligence based client server facial recognition attendance system engineered for deployment of old attendance system The system Uses a FastAPI backend performing 68 point facial landmark detection via dlib to generate 128 dimensional face embeddings and compared it through Euclidean distance thresholding. A React 18 frontend built with Vite delivers a glassmorphic dark themed interface with integrated 3D Spline visualizations. Real-time video is streamed as MJPEG over Local Host, and attendance records are logged directly to Google Sheets via the googlespread library. A supplementary Streamlit-based Student Portal deployed at vision-mark.in provides public read-only access to attendance statistics. The testing has done During Few lectures in classroom and it achieved 94% identification accuracy with sub 500ms per face latency reducing session start overhead to under 4-5 minute. The architecture prioritizes reproducibility and zero cost deployment suitable for resource constrained educational institutions.

Shayan Ahmad, Usaid Kaskar, Shoaib Shaikh, Roohan Shaikh, Ms.Samruddhi Santosh Kamble, Ali Karim Sayed,"VisionMark: A High-Performance Facial Recognition Attendance System" International Journal of Scientific Research and Engineering Development, V9(2): Page(1407-1413) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.