Crowd analysis is a widely studied automation problem with many surveillance applications, such as crowd monitoring, counting, public safety and security, and space design. Similarly, in autonomous driving, crowd detection, counting, and analysis are fundamental tasks. However, an essential aspect of these tasks is that the datasets are not suitable for detection-based approaches, as most crowd datasets suffer from perspective distortion. To overcome this issue, normalization is often required before analyzing the crowd. Recently, Unmanned Aerial Vehicles (UAVs) have become very popular for surveillance applications, offering a cost-effective solution and high-resolution images of crowded areas. The article benchmarks state-of-the-art object detection and density-based methods utilizing UAV aerial photographs on a new crowd-counting dataset called VisDrone. We aim to evaluate and compare various crowd analysis and counting techniques, explicitly focusing on using high-resolution UAV imagery to address the challenges associated with conventional ground-level camera datasets. Furthermore, we conduct extensive experiments to provide the results through commonly used metrics with accompanying insights, future direction, and conclusions. We hope this article will provide the community with baselines on the VisDrone dataset.