We have worked on quite a few models for crowd counting using drones and we do have a baseline pre-trained model for it.
However, we recommend that you still provide around 50 to 100 samples from your specific use-cases due to the differences that exist on a per drone basis (camera settings, altitude, geographic area/topography).

You will like one of our blog on using drone to detect objects - https://blog.nanonets.com/how-to-easily-do-object-detection-on-drone-imagery-using-deep-learning/

Did this answer your question?