<![CDATA[Attendance-data validity was essential to support accurate distribution of the Free Nutritious Meals Program. This study designed and implemented a Raspberry Pi 4-based hybrid attendance-verification system with selectable facial and fingerprint recognition. Experimental engineering was applied through device design, dataset preparation, software integration, and subsystem and integrated-system testing. Facial recognition used YOLOv5n as a trigger, MediaPipe for face detection, MobileFaceNet for embedding extraction, and cosine similarity for identity matching. The dataset produced 1,172 valid embeddings from 30 students. Testing achieved 100% fingerprint success, 100% face-detection success, and 94.7% facial-recognition accuracy. All five storage and web-dashboard functions operated as designed. The system supported local, automatic, and integrated attendance recording; however, facial-recognition performance decreased under low illumination.]]>
Copyrights © 2026
