Concrete is a fundamental material in infrastructure often exhibits significant surface cracking over time due to exposure to harsh environmental conditions. Cracking is considered one of the most critical forms of deterioration. In recent years, numerous disasters globally have underscored this vulnerability. Visual sensing and deep learning has marked a significant evolution, offering a leap from the manual, expert-dependent methods to a more continuous, data-driven approach. This improvement facilitated the integration of new components for structural health monitoring such YOLO series. It is characterized by their affordability, high resolution, and reliability. This paper aims to provide a short review of the current state of deep learning methodologies in real-time crack detection, with a particular focus on the application of the YOLO series. It presents also the main key components in the YOLO platform, the progresses and updates in each release. Despite the significant advancements to enhance detection capacities, several challenges remain. Summary of challenges and recommendations will be underlined at the end of this article.