PVD/modules/functional/sampling.py

90 lines
3.6 KiB
Python
Raw Normal View History

2021-10-19 20:54:46 +00:00
import numpy as np
import torch
from torch.autograd import Function
from modules.functional.backend import _backend
2023-04-11 09:12:58 +00:00
__all__ = ["gather", "furthest_point_sample", "logits_mask"]
2021-10-19 20:54:46 +00:00
class Gather(Function):
@staticmethod
def forward(ctx, features, indices):
"""
Gather
:param ctx:
:param features: features of points, FloatTensor[B, C, N]
:param indices: centers' indices in points, IntTensor[b, m]
:return:
centers_coords: coordinates of sampled centers, FloatTensor[B, C, M]
"""
features = features.contiguous()
indices = indices.int().contiguous()
ctx.save_for_backward(indices)
ctx.num_points = features.size(-1)
return _backend.gather_features_forward(features, indices)
@staticmethod
def backward(ctx, grad_output):
2023-04-11 09:12:58 +00:00
(indices,) = ctx.saved_tensors
2021-10-19 20:54:46 +00:00
grad_features = _backend.gather_features_backward(grad_output.contiguous(), indices, ctx.num_points)
return grad_features, None
gather = Gather.apply
def furthest_point_sample(coords, num_samples):
"""
Uses iterative furthest point sampling to select a set of npoint features that have the largest
minimum distance to the sampled point set
:param coords: coordinates of points, FloatTensor[B, 3, N]
:param num_samples: int, M
:return:
centers_coords: coordinates of sampled centers, FloatTensor[B, 3, M]
"""
coords = coords.contiguous()
indices = _backend.furthest_point_sampling(coords, num_samples)
return gather(coords, indices)
def logits_mask(coords, logits, num_points_per_object):
"""
Use logits to sample points
:param coords: coords of points, FloatTensor[B, 3, N]
:param logits: binary classification logits, FloatTensor[B, 2, N]
:param num_points_per_object: M, #points per object after masking, int
:return:
selected_coords: FloatTensor[B, 3, M]
masked_coords_mean: mean coords of selected points, FloatTensor[B, 3]
mask: mask to select points, BoolTensor[B, N]
"""
batch_size, _, num_points = coords.shape
2023-04-11 09:12:58 +00:00
mask = torch.lt(logits[:, 0, :], logits[:, 1, :]) # [B, N]
2021-10-19 20:54:46 +00:00
num_candidates = torch.sum(mask, dim=-1, keepdim=True) # [B, 1]
masked_coords = coords * mask.view(batch_size, 1, num_points) # [B, C, N]
2023-04-11 09:12:58 +00:00
masked_coords_mean = (
torch.sum(masked_coords, dim=-1) / torch.max(num_candidates, torch.ones_like(num_candidates)).float()
) # [B, C]
2021-10-19 20:54:46 +00:00
selected_indices = torch.zeros((batch_size, num_points_per_object), device=coords.device, dtype=torch.int32)
for i in range(batch_size):
current_mask = mask[i] # [N]
current_candidates = current_mask.nonzero().view(-1)
current_num_candidates = current_candidates.numel()
if current_num_candidates >= num_points_per_object:
choices = np.random.choice(current_num_candidates, num_points_per_object, replace=False)
selected_indices[i] = current_candidates[choices]
elif current_num_candidates > 0:
2023-04-11 09:12:58 +00:00
choices = np.concatenate(
[
np.arange(current_num_candidates).repeat(num_points_per_object // current_num_candidates),
np.random.choice(
current_num_candidates, num_points_per_object % current_num_candidates, replace=False
),
]
)
2021-10-19 20:54:46 +00:00
np.random.shuffle(choices)
selected_indices[i] = current_candidates[choices]
selected_coords = gather(masked_coords - masked_coords_mean.view(batch_size, -1, 1), selected_indices)
return selected_coords, masked_coords_mean, mask