Comparative Analysis of Object Detection Accuracy in Diverse Background Conditions

Object detection models must perform effectively across diverse environmental conditions to support autonomous navigation and surveillance systems. YOLOv5 and MobileNetV2 were evaluated using a custom dataset of 100,000 images captured under varied conditions, including daylight, fog, rain, and nighttime. YOLOv5 achieved an F1 score of 0.89 and a precision of 0.91 in low-light and rainy scenarios, demonstrating robustness in challenging environments. MobileNetV2 delivered faster inference speeds of up to 45 FPS on devices like the Raspberry Pi 5 and Google Coral Dev Board, with an F1 score of 0.82 in clear and foggy conditions, making it a strong candidate for resource-constrained deployments. The evaluation highlights YOLOv5's advanced feature extraction for high-accuracy detection and MobileNetV2's lightweight architecture for computational efficiency. These findings enable informed decision-making when selecting models for edge-based object detection applications, particularly in cost-sensitive and real-world scenarios.