alvaro-budria

It seems the servers are down again... :(

I am also encountering the same error. Help is much appreciated...

@shirsneh @yujvv @AlmaDev40 @pkudzia I put together a Colab that installs OpenPose and circumvents the issue. It's divided in two parts, only the first part deals with OpenPose, I think...

I can confirm that the workaround suggested by @petrgeiger-incieve seems to be working on my end. I am using `PyTorch` `1.12` btw.

> @cornerfarmer Hi and thank you for your reply. I ran this command, it did install h5py After this I ran blenderproc quickstart > > but this resulted in the...

Thx! I will check the new code and try to come up with something for the curvature loss.

Hi, I am unsure about the current computation of the Laplacian: `laplace = (points_d_sdf[..., 0::2] + points_d_sdf[..., 1::2] - 2 * sdf[..., None]) / (eps ** 2)` It assigns a...

Yes, that seems correct. The code was taking the absolute value of each partial derivative, instead of summing them and then taking the absolute value.

Sure, I can check if this new curvature penalty improves the results. ~~The linear warmup for the curvature loss weight is not implemented, and in the paper they mention that...

I tried the corrected Laplacian computation with a subset of DTU scan scenes. I could not observe any improvement in terms of PSNR. | PSNR | 24 | 37 |...