The application(exe) of Purdue's segmentation was not found

[jieeeeeeeeeee]

Hello, how can I get the application(exe) of Purdue's segmentation and reconstruction? The code is here: https://github.com/Kitware/Danesfield/blob/4c7f5115c8998aa7965e5f639315ed826220e946/tools/roof_geon_extraction.py#L85 https://github.com/Kitware/Danesfield/blob/4c7f5115c8998aa7965e5f639315ed826220e946/tools/roof_geon_extraction.py#L148

Is this step is necessary for building reconstruction？

[mleotta]

Good question. The Purdue code was written in C/C++, which is why it's not checked into main Python repository. In fact, I think we have not yet pushed that source code out on Github yet, we need to do that. However, the software is available in binary form as part of the conda environment.

https://github.com/Kitware/Danesfield/blob/master/deployment/conda/conda_env.yml#L50

All of this is pre-installed on a docker container if that's easier

https://hub.docker.com/r/kitware/danesfield

[jieeeeeeeeeee]

Thanks for your reply. I tried to run the entire pipeline under the windows platform. After hard work, I found the code of the 'segment' here. I almost completed all the steps, but in the end, I still failed because I didn't find the code of 'reconstruct'. I especially hope to get the code of 'reconstruct' so that I can run the whole pipeline under windows.

[mleotta]

We will try to get that code pushed to Github, but I don't know if it compiles on Windows. Your best be might be using Windows subsystem for Linux.

[jieeeeeeeeeee]

That is a good idea, I run the application of 'reconstruct', download from https://hub.docker.com/r/kitware/danesfield, in my subsystem of Linux, The out of the 'reconstruct' is a ply file namely 'building _0.ply'. This ply file can not be opened directly using MeshLab and CloudCompare: I looked into the ply file and find some specially defined geometry of 'Gabal Planes', 'Falts Planes, and 'Shelds Planes'. The ply like this:

I want to know if this file is normal? and what software can I use to directly open and visualize it? Thinks.

[jieeeeeeeeeee]

I deleted the specially defined geometry in ply file and opened it with MeshLab successfully. I got this result:

I think it does not like a building but a lot of planes.

[mleotta]

I've not had issues opening the PLY files in meshlab before, but this output might not be the final output file and I usually open those final outputs. It might be a later processing step that extrudes roof planes to form building walls, and then another later step that applies texture.

Is this the Jacksonville data? I recognize a hint of that U-shaped building, but this does not look like the typical output of the processing pipeline. Did you run the whole pipeline or just up to that reconstruct step?