Object Detection in Low-Light Conditions: A Comparison using YOLOv5 and YOLOv8

Object detection refers to identifying objects by determining their presence, location, and size, and assigning appropriate class labels. Detecting objects under low-light conditions presents challenges due to insufficient illumination, leading to decreased object visibility. In previous experiments, YOLOv5 and YOLOv8 were used to detect objects. This study aims to evaluate the effectiveness of the YOLOv5 and YOLOv8 models in detecting objects in low-light scenarios. The datasets used in this research contain 4,132 low-light images obtained from the publicly available Roboflow dataset. Evaluating the performance of YOLOv5 and YOLOv8 models involves using metrics such as precision, recall, F1-score, and mAP values. Based on a comparison performance, YOLOv8 outperforms YOLOv5. The results of precision, recall, F1-Score, and mAP values indicate that the YOLOv8 model performance exceeds the YOLOv5 model, with differences of 6.14%, 10.61%, 9%, and 11.89%, respectively, for each performance metric.