<![CDATA[Limited computing resources and high hardware costs often limit the implementation of facial recognition systems on embedded devices. This study aims to evaluate and compare the performance of Support Vector Machine (SVM) and Convolutional Neural Network (CNN) algorithms on a Raspberry Pi 4B microcomputer. The research method involves testing the CNN architecture and SVM kernel on a Kaggle dataset consisting of 2,000 facial images from five identity classes. The evaluation parameters used include accuracy, precision, recall, F-Score, resource usage (CPU/RAM), and inference speed. The test results show that the CNN algorithm achieves 93% accuracy but takes longer inference time, averaging 268.52 ms per image. On the other hand, SVM achieves 87% accuracy with much faster inference time, averaging 8.02 ms per image. Based on the test results, this study concludes that although CNN is superior in accuracy, SVM is more recommended for real-time biometric system applications on microcomputers due to its computational time efficiency and lower resource usage.]]>
Copyrights © 2026
