How about training point clouds with different size?

[leihouyeung]

Hi, thanks for this fantastic work. In the paper, you said "The methods we describe here extend easily to the M = N case
because DGCNN, Transformer, and Softmax treat inputs as unordered sets. None requires X and Y to have the same length or a bijective matching."

Is there any implementation of the cases when M differs from N? Thanks.

[luisfmnunes]

Usually what is done in the N x M case is to pad entries to a common value (example K > M > N) and then use the mask component of the Attention Module to neglect the effect of padded data.