Utilizing Convolutional Neural Networks to Classify Dry Road Surface Conditions

The abstract gives the performance metrics of a model developed to estimate dry road surface conditions from impact ratings. The model's high precision and recall on all impact ratings were marvelous, but there were also Impact Rating 1 at a precision of 98.56% and a recall of 98.57%, Impact Rating 2 at 98.52%, Impact Rating 3 at 98.56%, Impact Rating 4 at 98.54%, and Impact Rating 5 equally at 98.54%. With 10,350 photos in the dataset, the model's overall accuracy was 98.55%. These values are also reflected in the F1scores clearly showing that the model had balanced performance within each class. The model further had a marvelous accuracy about all classes in total, standing at 99%. Support values that signify the number of samples for each impact rating ranged from 2,030 to 2,100 samples and were evenly apportioned to support every class to the proportion of 0.20. The consistency strengthens the model's robustness and reliability in its decision to classify dry road surfaces. The combination of the above metrics indicates that the model has the strong potential to judge these grades of road conditions properly, thus having promising applicability in numerous road maintenance, safety assessment, and management infrastructures. In short, high results from the model outcome indicate that the model would assist correctly in improving the strategies to enhance road safety and maintenance by eventually providing high-quality evaluations for roads.