Aircraft designed for transporting one or two persons, much like a small personal car but flying in the air based on Bernoulli Principle and Newton's laws, are called Personal Aerial Vehicles (PAVs). Due to high PAV traffic densities and the high velocities at which PAVs fly, manual piloting of PAVs is seldom recommended. Hence, PAVs are equipped with an inbuilt Autonomous Navigation and Control System (ANCS), which frees the rider from piloting skills. Machine learning (ML)-based approaches that require datasets used during training phases can be used for implementing the software of such ANCS. The development of simulation-driven systems for ANCS offers many advantages, particularly reducing the systems' developmental costs and infrastructure needs. In this article, we report on the development of a synthetic visual dataset that enables ML-based implementation of ANCS. The state-of-the-art simulator AirSim is used to generate this dataset. Additionally, to make the synthetic data more realistic, several augmenting technologies such as Unreal Engine (3D graphics gaming), Blender animator, PX4-Autopilot SITL (Software in the loop) software, QGroundControl Autopilot App, and ROS (robot operating system) middleware suite are utilized. We also discuss the applicability of this dataset in realizing the ANCS module for PAVs.