29 lines
1,015 B
Python
29 lines
1,015 B
Python
|
import torch
|
||
|
|
import torch.nn as nn
|
||
|
|
|
||
|
|
import modules.functional as F
|
||
|
|
|
||
|
|
__all__ = ['Voxelization']
|
||
|
|
|
||
|
|
|
||
|
|
class Voxelization(nn.Module):
|
||
|
|
def __init__(self, resolution, normalize=True, eps=0):
|
||
|
|
super().__init__()
|
||
|
|
self.r = int(resolution)
|
||
|
|
self.normalize = normalize
|
||
|
|
self.eps = eps
|
||
|
|
|
||
|
|
def forward(self, features, coords):
|
||
|
|
coords = coords.detach()
|
||
|
|
norm_coords = coords - coords.mean(2, keepdim=True)
|
||
|
|
if self.normalize:
|
||
|
|
norm_coords = norm_coords / (norm_coords.norm(dim=1, keepdim=True).max(dim=2, keepdim=True).values * 2.0 + self.eps) + 0.5
|
||
|
|
else:
|
||
|
|
norm_coords = (norm_coords + 1) / 2.0
|
||
|
|
norm_coords = torch.clamp(norm_coords * self.r, 0, self.r - 1)
|
||
|
|
vox_coords = torch.round(norm_coords).to(torch.int32)
|
||
|
|
return F.avg_voxelize(features, vox_coords, self.r), norm_coords
|
||
|
|
|
||
|
|
def extra_repr(self):
|
||
|
|
return 'resolution={}{}'.format(self.r, ', normalized eps = {}'.format(self.eps) if self.normalize else '')
|