Regarding downsampling layers in ounet and unet

[raoshashank]

I notice that the downsampling layers in the ounet and unet models use ConvBnRelu as the downsamplers rather than pooling, which is typically used in such dense prediction/reconstruction networks to extract hierarchical features. Is there some relation between the two operations due to the structure of the octree itself?

[wang-ps]

The stride of ConvBnRelu is 2 for downsampling; a pooling layer with a stride of 2 can also be used for downsampling. The ConvBnRelu has trainable parameters, it can result in slightly better preformance than pooling.