The availability of diverse and comprehensive datasets is crucial for training and testing Artificial Intelligence (AI) algorithms. However, traditional data collection methods are often costly, time-intensive, and raise concerns regarding security and privacy. This is particularly evident in domains such as automotive and smart mobility sectors, where acquiring real-world data is both expensive and logistically complex. To address these challenges, this paper explores the use of synthetic data generation through deep generative models. We focus on both unconditional and conditional approaches to image synthesis using domain-specific data from the city of Hanover, Germany. We utilize different architectures, and specifically investigate the challenges related to data security, diversity, and scalability in these models. Our evaluation assesses model performance across image quality, sample diversity, and convergence speed under data-constrained conditions. The results demonstrate the potential of synthetic data to enhance dataset diversity and support AI development in real-world urban environments, offering practical insights for industrial applications such as autonomous driving and field data enrichment.