The November 2020 solution release sees the first integration of deep learning into the 3D basemaps solution. Leveraging a new tree classification deep learning model on the Living Atlas, it is now possible to classify trees in lidar point clouds in ArcGIS Pro.
Before and after using deep learning to classify trees in the lidar point cloud
This solution release also adds a new density based cluster approach to tree extraction which significantly improves the tree extraction compared to a raster based approach.
Extracting tree location, height and width using density based clustering.
Just keep in mind that density based clustering on large amount of points is much more resource intensive than the existing raster based approach so we recommend using the latter first to get a generic tree coverage and use the density based cluster approach only for areas where you need accurate tree extraction.
For more information, see the 3D basemap solution documentation.