Analisis Perbandingan Metode Yolo Dan Faster R-CNN Dalam Deteksi Objek Manusia

Abstract

Human object detection is an important component in surveillance systems, behavior analysis, and crowd management in public spaces such as stadiums, shopping malls, and terminals. However, the detection process often faces obstacles such as inconsistent lighting, complex backgrounds, and high object density. This study aims to compare the performance of two object detection algorithms, namely YOLOv10 and Faster R-CNN, in detecting humans. The dataset used is uniform and covers a wide range of environmental conditions to ensure fair and objective evaluation. This research involves the stages of data collection, pre-processing, model training, testing, and performance evaluation. The test results show that YOLOv10 has a performance advantage with an mAP50 value of 0.75, higher than that of Faster R-CNN which obtained an AP50 of 0.67. Based on these findings, YOLOv10 is recommended for use in applications that require real-time human detection with a high level of accuracy. Kata kunci: YOLOV10; Faster R-CNN; Object Detection