Many cities and road management organizations are motivated to automate the assessment of road damage to enhance safety for both self-driving vehicles and general road users. This research addresses the challenges of limited access to advanced equipment and expertise by utilizing the diverse road damage dataset (RDD2022) from six countries, comprising over 55,000 instances of various road damage types. Here a hybrid approach is proposed that integrates deep learning with fuzzy logic for the automatic detection, classification, and severity assessment of road damage. The methodology includes several key steps: data preprocessing, deep learning model training using a ResN et50 architecture, fuzzy inference system design, and system integration. The convolutional neural network (CNN) processes road damage images to classify different types, such as longitudinal cracks and potholes, while the fuzzy logic system evaluates the severity based on inputs like damage size and position derived from the CNN's outputs. The proposed method reaches an accuracy of about 95.26%, comparable to the existing methods. The proposed solution could help road management organizations to set up intelligent systems for automatic monitoring of road conditions.