This paper aims to develop a deep learning-based object detection framework for the intelligent recognition of road diseases. By thoroughly analyzing deep learning technologies and their application in road condition detection, the YOLO algorithm was selected and optimized. Considering the existence of ineffective stacking in convolutional layers, a Transformer model was introduced into the algorithm to fully extract feature information. The Global Road Damage Detection Challenge 2020 dataset was used for training and testing, and model performance was evaluated using metrics such as the F1 score, which is the weighted average of recall rates. The results indicate that the optimized YOLO model achieves an accuracy of approximately 60% in road condition detection, showing a significant improvement in efficiency compared to traditional methods.