Surveillance is critical for ensuring safety and security in smart cities, and efficient systems for monitoring urban spaces are essential. Traditional surveillance systems overly rely on human labor and frequent maintenance. However, artificial vision techniques reduce such dependence by automatically analyzing the information in the context of intelligent surveillance using aerial imagery. In this study, we assess some state-of-the-art (SOTA) models for general object detection that excel when using high-quality imagery in urban surveillance settings. We then analyze their performance using the Dataset of Object deTection of Aerial Images (DOTA) standard benchmark to determine which models fit the specific demands of aerial surveillance. Thus, our experimental results reveal valuable insights for city planners, security agencies, and public safety organizations looking to enhance their surveillance capabilities through AI-driven technologies prioritizing real-world scenarios over ideal conditions.