Does the training point cloud data of PointSO include normals?

[anon-mity]

I believe your work makes a valuable contribution to the community, and your paper is highly insightful. However, I have a question to ask: Regarding the PointSO model within the SOFAR framework, I noticed that the input channel count of the first layer in its point cloud encoder is 6 instead of 3. Does this indicate that you additionally used normal information? Yet, in the Dataset code, I did not see unified rotation applied to the normals—it seems that only the point cloud positions were rotated.

Chinese: 对于SOFAR框架中的PointSO模型，我注意到点云编码器第一层的输入通道数是6而不是3，这是否代表着你们还额外使用了法线信息？但我在DataSet的代码中并没有看到对法线的统一rotation（似乎只对点云位置进行了旋转）？

[qizekun]

Hello, we use colored point clouds, i.e., xyzrgb