search

tudelft3d / City3D

Large-scale LoD2 Building Reconstruction from Airborne LiDAR Point Clouds
https://github.com/tudelft3d/City3D
GNU General Public License v3.0
237 stars 39 forks source link

The problem with the results in Vaihingen data #19

Closed lihanhan closed 1 year ago

lihanhan commented 1 year ago

Hello, I am sorry to disturb you. I conducted the experiment without adding the footprint on the vaihingen data. However, the results are very different from the results of the paper. Is there a parameter or a specific step that needs to be changed?Because even if I change the alpha parameter in the alphashape process, I still can't get a better result. I would like to get your help !

image

yidahuang commented 1 year ago

In my paper, we use a training set of Vaihingen dataset that contains footprint, why do you need to generate footprint?

lihanhan commented 1 year ago

Ok, thanks for your reply. I would like to study the algorithm for generating footprint to get a good architectural model. Is the building footprint file in DXF format in the training set containing the building footprint that you mentioned?

lihanhan commented 1 year ago

Ok, thanks for your reply. I would like to study the algorithm for generating footprint to get a good architectural model. Is the building footprint file in DXF format in the training set containing the building footprint that you mentioned?

yidahuang commented 1 year ago

From my point of view, footprint generation may follow this pipeline: 3D point projection-->boundary extraction(using alpha shape or other)--> regularization and simplification (involving the detection of main direction, potential orthogonalization, parallelism, etc.). Hope it helps (: Yes, the .dxf file is what you want, you can directly visualize it using qgis or other software.

lihanhan commented 1 year ago

Thanks for your reply and good luck with your work!