My goal is to do a rough treetop calculation to get an idea of how many trees are there and how they are layouted.
For such goal, it is not necessary to remotely identify individual trees. A simpler approach to get an idea of tree density would be to model it with some metric related to canopy cover. For example, using the NDVI index to map which pixels belong to vegetation (example), and then, quantify these pixels.
A limitation is that it would be necessary to get data (tree density) from field-derived plots to calibrate such model. Another caveat is that in dense forests (> 1000 trees/hectare) the canopy cover metric would start saturating and failing to capture the actual variation.
I believe this approach is more valid if the study area is large. If it is small then, acquiring a high-resolution image and digitizing the trees for counting might be a good idea (as suggested by CBG).