Jhonve
Release the training data generation procedure?
Hi, thanks for sharing your work.
Currently, I noticed that you had released your original complete mesh data, but the GCN-Denoiser took local patches for training, and the local path generation procedure is still missing.
Could you please release the complete procedure for the generation training patch for running your training code?
The local patch generation procedure example can be found in this line. Note that if you want to re-train our model to achieve the reported score, please follow the data balance strategy which means you have to add face class classification after tensor voting somewhere in this part.
could you provide a bit more detail for usage? something like the steps laid out in this repo: https://github.com/nini-lxz/DNF-Net
It's not clear to me how to use your codebase to generate the patches before training (and construct after inference).
