I am using a SSH connection. How should I modify the code？

[whu-lee]

I am using a SSH connection, so I not have a display to generate the images with Qt?
How should I modify the code?

If I comment self.draw_mesh(),

in viewGenerator.py
buffer = glReadPixels(0, 0, self.opts["imsize"], self.opts["imsize"], GL_RGB, GL_UNSIGNED_BYTE) im = Image.frombytes(mode="RGB", size=(self.opts["imsize"], self.opts["imsize"]), data=buffer)
This im will become all 0 ！
Kindly let me know how to resolve this problem.

[aboulch]

Hello,
This is a known issue of the current code, it requires to open a window.
You may try a 'ssh -X'.
If you are able to find a way to dowithout window opening, tell me.

[whu-lee]

Thank you for your reply.
Because I have just started to get involved in point cloud segmentation, I am sorry that I have no good ideas about this issue.
In addition, for the trained model of the pytorch version, can you share it?

[whu-lee]

The semantizer is not defined in line 53 of sem3d_test_to_labels.py. I am using the pytorch framework. What should I change the semantizer to?
I tried to change to

Sem3d.mesh_to_label_file_no_labels(mesh_filename,sem3d_cloud_txt,output_results)

However, it will report the following error:
AttributeError: module'semantic3D_utils.lib.python.semantic3D' has no attribute'mesh_to_label_file_no_labels'

At the beginning I have already Building To generate the C++/Python library

cd semantic3D_utils
python setup.py install --home="."