Thanks for providing the implementation of PolyGen. I find that the sampling strategy implemented in the code (for both the vertex model and the face model) is not exactly the same as described in the paper. The latter would mask out invalid samples, while the implemented one would not. I think that implementing such logic in Tensorflow is not very easy for me due to many template-related bugs I would always encounter.
Could you kindly provide the sampling code as described in the paper?
Hi, @saran-t
Thanks for providing the implementation of PolyGen. I find that the sampling strategy implemented in the code (for both the vertex model and the face model) is not exactly the same as described in the paper. The latter would mask out invalid samples, while the implemented one would not. I think that implementing such logic in Tensorflow is not very easy for me due to many template-related bugs I would always encounter.
Could you kindly provide the sampling code as described in the paper?
Thanks a lot!