Comparative Landslide Hazard Detection Analysis Based on ResNet and CNN

Landslide hazard prediction is crucial for disaster prevention and risk management. The rapid development of deep learning technology provides a low-cost and efficient detection method for landslide detection, but in the field of landslide hazard detection, there are few articles that use multiple neural network models simultaneously for landslides for comparative detection. This study aims to investigate the effectiveness of Convolutional Neural Networks (CNN) and Deep Residual Networks (ResNet) in landslide hazard detection. This design proposes three deep learning models for comparative detection of landslide hazards, which are 12-layer Convolutional Neural Networks (CNN-12) and ResNet18 and ResNet50 models, which are the more classical of ResNet network models. It is found that all three models have good detection results, with the ResNet18 model performing best on the validation set, with 99.95% recognition accuracy, F1 score, recall, and precision of 1. The self-designed CNN-12 model has the next best performance, and the ResNet50 has the worst performance.