Efficient and accurate detection of road surface cracks is crucial for ensuring road safety and reducing maintenance costs. To address this problem, the current paper proposes a new autonomous crack detection framework that involves the modified ResNet DLM. The proposed approach is a significant shift from the current manual scanning techniques that are prone to reliability and consistency. By doing so, the power of deep learning, particularly using the deep convolutional neural network system, has been established. The model is substantially more accurate and reliable in terms of the number of features of roads tainted for imperfection. The findings based on the current experiment have shown that the modified ResNet produced a 99.93% accuracy percentage. Other models, including VGG16, DenseNet, and MobileNet, were inferior to the modified ResNet. This study offers a glimpse of the future and the potential of learning to transform monitoring road quality by offering the hope for significant advances across the various fronts of infrastructure and safety guidelines.