Dear authors,
Thank you for releasing DualBrep. I tested the model using the test sample 00983750.ply by scaling the shape along the Y-axis with factors of 0.5, 0.75, 1.25, 1.5, and 2, while keeping all other conditions unchanged.
00983750.zip
The results show that only the versions scaled by 0.75 and 1.25 could be reconstructed as watertight, closed solids. The versions scaled by 0.5, 1.5, and 2 could not form watertight, closed solids. In addition, the overall quality of the reconstruction candidates was not satisfactory for any of the scaling factors.
Could you please explain why DualBrep is so sensitive to dimensional changes along a single axis? Is this mainly related to the training-data distribution, the input normalization method, or the post-processing stage? Do you have any recommended preprocessing methods or parameter settings for shapes with this type of non-uniform scaling?
Thank you for your help!
Dear authors,
Thank you for releasing DualBrep. I tested the model using the test sample 00983750.ply by scaling the shape along the Y-axis with factors of 0.5, 0.75, 1.25, 1.5, and 2, while keeping all other conditions unchanged.
00983750.zip
The results show that only the versions scaled by 0.75 and 1.25 could be reconstructed as watertight, closed solids. The versions scaled by 0.5, 1.5, and 2 could not form watertight, closed solids. In addition, the overall quality of the reconstruction candidates was not satisfactory for any of the scaling factors.
Could you please explain why DualBrep is so sensitive to dimensional changes along a single axis? Is this mainly related to the training-data distribution, the input normalization method, or the post-processing stage? Do you have any recommended preprocessing methods or parameter settings for shapes with this type of non-uniform scaling?
Thank you for your help!