Road Pothole Detection Model Based on AlexNet Convolutional Neural Network and Confusion Matrix

This paper presents a road pothole detection model based on the AlexNet Convolutional Neural Network (CNN). Through data preprocessing, model construction, Dropout regularization, and evaluation based on the confusion matrix, the model achieves efficient identification of road potholes. Data preprocessing includes cleaning, normalization, and augmentation to enhance the model's generalization capability. The AlexNet model extracts image features through multiple convolutional and pooling layers and classifies them using fully connected layers. Dropout technique randomly discards some neurons during the training process to effectively prevent overfitting. Model evaluation employs a confusion matrix and various classification metrics, including precision, accuracy, recall, specificity, F1 score, and AUC value. Experimental results show that the model exhibits an accuracy rate of over 93% and high generalization capability across different datasets, demonstrating the effectiveness and practicality of the model in road pothole detection.