search

samxuxiang / BrepGen

[SIGGRAPH 2024] Official PyTorch Implementation of "BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometry".
Other
227 stars 28 forks source link

Question about how to deal with 'boxx_scale' in sample.py #16

Open wangyian-me opened 3 months ago

wangyian-me commented 3 months ago

Noticed that when loading data to train EdgeZ diffusion, it rescale the bboxs and the corner points.

surf_pos = surf_pos * self.bbox_scaled
edge_pos = edge_pos * self.bbox_scaled 
corner_wcs = corner_wcs * self.bbox_scaled

And when sampling, it scales back the surf_pos and edge_pos by 

edge_pos = edgePos.detach().cpu().numpy() / 3.0
surfPos = surfPos.detach().cpu().numpy()  / 3.0

but why doesn't it also scale back the corner_wcs? Which is right here.