Comparative Analysis of YOLOv8 and Faster R-CNN for Real-Time Visual Anomaly Detection
Abstract
Automated anomaly detection in video streams is essential for industrial process monitoring, security surveillance, and quality control systems. This paper presents a systematic comparative analysis of two state-of-the-art deep learning object detectors, YOLOv8 and Faster R-CNN, applied to the task of visual anomaly detection. Using a dataset of 9,000 annotated frames containing four object classes including normal and anomalous patterns, we evaluate both architectures on detection accuracy, classification precision, and inference speed. Experimental results show that YOLOv8 achieves 85.4% precision and 86.6% recall at 30-40 frames per second on GPU, making it suitable for real-time applications. Faster R-CNN achieves superior accuracy (89.2% precision, 88.1% recall) but operates at only 8-12 frames per second. Confusion matrix analysis reveals that both detectors struggle with rare classes, achieving per-class F1 scores of 61-74% for minority categories compared to 91% for the majority class. ROC-AUC values are 0.89 for YOLOv8 and 0.94 for Faster R-CNN. Practical recommendations for detector selection based on application requirements are provided.