Hi, I’m new to Pytorch-geometric, I want to do mesh refining using graph cnn’s. I have an initial coarse mesh generated from a point cloud where the number of neighbours of each node is not same.I would like update the vertex(node) positions using graph network. I would like to know which type of graph convolution or graph network would be better for my experiment.