elementwise convolutionwith non-linear operation. e.g. per-pixel, per-point, per-voxel, per-facet, etc. Args: in_channels: numbe

This dataset suits meshes rendered offline. All information is stored in the .json file

Compute fuzzy coefficients for the facet2vertex convolution. We use spherical harmonics to achieve this. Returns: fuzzy filter coeffi

convolution to learn vertex features from adjacent facets Args: in_channels: number of input channels out_channels: number o

feature propagation from vertex to facet Args: in_channels: number of input channels out_channels: number of output channels

Compute fuzzy coefficients for the facet2vertex convolution. We use spherical harmonics to achieve this. Returns: fuzzy filter coeffi

convolution on the facet textures Args: in_channels: number of input channels out_channels: number of output channels use_

3D separable convolution with non-linear operation. Args: in_channels: number of input channels out_channels: number of output

Compute fuzzy coefficients for the facet2vertex convolution Returns: fuzzy filter coefficients

3D Global Mesh pooling.

elementwise convolution with non-linear operation. e.g. per-pixel, per-point, per-voxel, per-facet, etc. Args: in_channels: numb

3D Mesh pooling. Args: pooling method, default to 'max' Returns: Variable tensor

3D point cloud pooling. Args: pooling method, default to 'max' Returns: Variable tensor of shape [concat_Mp,C]

3D Mesh unpooling Returns: Variable tensor

3D unpooling Returns: outputs: float32 tensor of shape [concat_Np,C], unpooled features