Reconstruction of custom dataset

[julesmorel]

Describe the bug
I do not manage to reconstruct buildings on a custom dataset.

• I converted my custom dataset into NumPy binary files using [points2surf/make_pc_dataset.py](https://github.com/ErlerPhilipp/points2surf/blob/master/make_pc_dataset.py).
• Then I extracted the planar primitives from point clouds with Mapple and saved the.vgfiles into into ./datasets/custom_dataset/06_vertex_group

However, when i run python reconstruct.py dataset_name='custom_dataset' model_name='helsinki_fullview', the program ends on :

[2024-09-10 18:10:20,215][root][INFO] - cut performed: 0.00 s
[2024-09-10 18:10:20,215][root][INFO] - cut_value: 0.00
[2024-09-10 18:10:20,215][root][INFO] - number of extracted cells: 0
[2024-09-10 18:10:20,215][root][ERROR] - no reachable cells. aborting
[2024-09-10 18:10:20,622][root][INFO] - cut performed: 0.00 s
[2024-09-10 18:10:20,622][root][INFO] - cut_value: 0.00
[2024-09-10 18:10:20,622][root][INFO] - number of extracted cells: 0
[2024-09-10 18:10:20,622][root][ERROR] - no reachable cells. aborting
[2024-09-10 18:10:21,833][root][INFO] - cut performed: 0.00 s
[2024-09-10 18:10:21,833][root][INFO] - cut_value: 0.00
[2024-09-10 18:10:21,833][root][INFO] - number of extracted cells: 0
[2024-09-10 18:10:21,833][root][ERROR] - no reachable cells. aborting
[2024-09-10 18:10:21,852][root][INFO] - cut performed: 0.00 s
[2024-09-10 18:10:21,853][root][INFO] - cut_value: 0.00
[2024-09-10 18:10:21,853][root][INFO] - number of extracted cells: 0
[2024-09-10 18:10:21,853][root][ERROR] - no reachable cells. aborting

...

The installation seems to be good because I manage to produce the .obj files for the demo dataset helsinki_mini.
I guess it is not a bug but rather a parameter tuning problem.
Any input to solve this would be greatly appreciated.

Environment
Ubuntu 22.04 WSL

[chenzhaiyu]

Hi, this is most likely because your point cloud data is very different from the data for training the model. See also #18.

[chl916185]

`import numpy as np
import os
import random
import open3d as o3d
ss_path = 'helsinki_mini/helsinki_mini/04_pts/1.xyz.npy'
xyz = np.load(ss_path)
xyz = np.array(xyz)
x_min = xyz[:,0].min()
y_min = xyz[:,1].min()
z_min = xyz[:,2].min()

xyz[:,0] -= x_min
xyz[:,1] -= y_min
xyz[:,2] -= z_min

xyz +=1

xyz *=100
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(xyz)

o3d.io.write_point_cloud('real_world/00_base_pc/my_1.ply',pcd,write_ascii=True)
print()
`
我用示例数据，做了缩放处理，得不到结果是什么原因呢
@chenzhaiyu