Evaluating the Performance of ResNet-50 and GoogleNet for Damage Detection and Classification

Accurate and rapid assessment of structural damage is critical for maintaining the safety and integrity of infrastructure, especially following natural disasters. Traditional methods of damage assessment, which rely on manual inspections, can be labor-intensive and subject to human error. This study exam-ines the use of ResNet50 and GoogLeNet convolutional neural networks (CNNs) to automate the classification of damaged structures. The models were trained on a diverse dataset that includes images of both damaged and undamaged structures, ensuring their adaptability to different conditions. Evaluation metrics such as precision, recall, accuracy, and F1-score were used to measure performance. The results demonstrate that both ResNet50 and GoogLeNet are effective for assessing damage in post-disaster contexts. However, GoogLeNet shows a slight advantage achieving testing accuracy of 97.5% as compared to ResNet50 achieving 97.2% accuracy. This suggests that GoogLeNet offer a more reliable option for real-world disaster response applications.