search

tudelft3d / PSSNet

PSSNet: Planarity-sensible Semantic Segmentation of Large-scale Urban Meshes
GNU General Public License v3.0
36 stars 6 forks source link

which data is the "mesh over-segmentation results" and "graphs" in the step-2? #1

Closed titlezi closed 1 year ago

titlezi commented 1 year ago

Hi author: your work is really nice! In the step-2, I don't understand which data is "mesh over-segmentation results" , I try to use segments_pnp and it failed. Does "mesh over-segmentation results" mean the feature_pnp, and "graphs" mean spg_inputs/graphs_edges?

Yours! Best regards! Tian

WeixiaoGao commented 1 year ago

Hi Tian,

Thank you for your interest in our project. We generate mesh over-segmentation results only in step-1. You can simply run PSSNet_pipeline_for_GCN in step-1 to generate all data for step-2. In step-2, we do not do the mesh over-segmentation, but we perform the GCN-based semantic segmentation. The input for step-2 should be in the folder spg_input. You can simply copy the data in spg_input/pcl/ to ../datasets/custom_set/data/ and the data in spg_input/graph_edges/ to ../datasets/custom_set/pssnet_graphs/. The data in segments_pnp is only for visualization. The mesh over-segmentation results for step-2 are encoded in the data in spg_input/pcl/, and the graph is encoded in the data in spg_input/graph_edges/. Int step-2, we just use the pssnet_partition.py to convert from .ply to .h5 for later processing in pssnet_main.py.

Kind regards, Gao

titlezi commented 1 year ago

Hi Tian,

Thank you for your interest in our project. We generate mesh over-segmentation results only in step-1. You can simply run PSSNet_pipeline_for_GCN in step-1 to generate all data for step-2. In step-2, we do not do the mesh over-segmentation, but we perform the GCN-based semantic segmentation. The input for step-2 should be in the folder spg_input. You can simply copy the data in spg_input/pcl/ to ../datasets/custom_set/data/ and the data in spg_input/graph_edges/ to ../datasets/custom_set/pssnet_graphs/. The data in segments_pnp is only for visualization. The mesh over-segmentation results for step-2 are encoded in the data in spg_input/pcl/, and the graph is encoded in the data in spg_input/graph_edges/. Int step-2, we just use the pssnet_partition.py to convert from .ply to .h5 for later processing in pssnet_main.py.

Kind regards, Gao

Thanks for your reply. Best regards! Tian