Implementation: Python 3.7

'packages' segmentation_func: useful functions for the projection, detection, segmentation and reverse projection pipeline alienlab: useful functions for image display and saving

'content' Detec_Segm_0: Pipeline described in the reference article (with slight modifications) Detec_Segm_Loop: Supervised segmentation of trees using class ground truth

Change the file_path in the programs for other point clouds (here it works automatically for file_path = 'Cassette_idclass/Cassette_GT.ply' (Rue Soufflot point cloud))

Remark: -the function to get the elevation is quite slow and could be speed up using numpy built in functions like np.unique

-If it doesn't work properly on other point clouds: Parameters that could be tunable: -kx and ky voxel parameter size If ground search fails: -Initial seed selection for the ground search (in segmentation_func, lambdaflat2) -Elevation criterion ground search -Neighbourhood size ground search If the computation of the performances fails: -check how are named the ground truth fields in the point cloud