-
Notifications
You must be signed in to change notification settings - Fork 23
/
scene_box.py
126 lines (106 loc) · 4.61 KB
/
scene_box.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
# Copyright 2024 the authors of NeuRAD and contributors.
# Copyright 2022 the Regents of the University of California, Nerfstudio Team and contributors. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Dataset input structures.
"""
from dataclasses import dataclass
from typing import Tuple, Union
import torch
import viser.transforms as vtf
from jaxtyping import Float
from torch import Tensor
@dataclass
class SceneBox:
"""Data to represent the scene box."""
aabb: Float[Tensor, "2 3"]
"""aabb: axis-aligned bounding box.
aabb[0] is the minimum (x,y,z) point.
aabb[1] is the maximum (x,y,z) point."""
def within(self, pts: Float[Tensor, "n 3"]):
"""Returns a boolean mask indicating whether each point is within the box."""
return torch.all(pts > self.aabb[0], dim=-1) & torch.all(pts < self.aabb[1], dim=-1)
def get_diagonal_length(self):
"""Returns the longest diagonal length."""
diff = self.aabb[1] - self.aabb[0]
length = torch.sqrt((diff**2).sum() + 1e-20)
return length
def get_center(self):
"""Returns the center of the box."""
diff = self.aabb[1] - self.aabb[0]
return self.aabb[0] + diff / 2.0
def get_centered_and_scaled_scene_box(self, scale_factor: Union[float, torch.Tensor] = 1.0):
"""Returns a new box that has been shifted and rescaled to be centered
about the origin.
Args:
scale_factor: How much to scale the camera origins by.
"""
return SceneBox(aabb=(self.aabb - self.get_center()) * scale_factor)
@staticmethod
def get_normalized_positions(
positions: Float[Tensor, "*batch 3"], aabb: Float[Tensor, "* 2 3"], per_dim_norm: bool = True
):
"""Return normalized positions in range [0, 1] based on the aabb axis-aligned bounding box.
Args:
positions: the xyz positions
aabb: the axis-aligned bounding box
"""
batch_shape = positions.shape[:-1]
if len(aabb.shape) > 2 and batch_shape != aabb.shape[:-2]:
# add singleton dimension
aabb = aabb.unsqueeze(1)
aabb_lengths = aabb[..., 1, :] - aabb[..., 0, :]
aabb_lengths = aabb_lengths if per_dim_norm else aabb_lengths.max(dim=-1, keepdim=True)[0]
normalized_positions = (positions - aabb[..., 0, :]) / aabb_lengths
return normalized_positions
@staticmethod
def from_camera_poses(poses: Float[Tensor, "*batch 3 4"], scale_factor: float) -> "SceneBox":
"""Returns the instance of SceneBox that fully envelopes a set of poses
Args:
poses: tensor of camera pose matrices
scale_factor: How much to scale the camera origins by.
"""
xyzs = poses[..., :3, -1]
aabb = torch.stack([torch.min(xyzs, dim=0)[0], torch.max(xyzs, dim=0)[0]])
return SceneBox(aabb=aabb * scale_factor)
@dataclass
class OrientedBox:
R: Float[Tensor, "3 3"]
"""R: rotation matrix."""
T: Float[Tensor, "3"]
"""T: translation vector."""
S: Float[Tensor, "3"]
"""S: scale vector."""
def within(self, pts: Float[Tensor, "n 3"]):
"""Returns a boolean mask indicating whether each point is within the box."""
R, T, S = self.R, self.T, self.S.to(pts)
H = torch.eye(4, device=pts.device, dtype=pts.dtype)
H[:3, :3] = R
H[:3, 3] = T
H_world2bbox = torch.inverse(H)
pts = torch.cat((pts, torch.ones_like(pts[..., :1])), dim=-1)
pts = torch.matmul(H_world2bbox, pts.T).T[..., :3]
comp_l = torch.tensor(-S / 2)
comp_m = torch.tensor(S / 2)
mask = torch.all(torch.concat([pts > comp_l, pts < comp_m], dim=-1), dim=-1)
return mask
@staticmethod
def from_params(
pos: Tuple[float, float, float], rpy: Tuple[float, float, float], scale: Tuple[float, float, float]
):
"""Construct a box from position, rotation, and scale parameters."""
R = torch.tensor(vtf.SO3.from_rpy_radians(rpy[0], rpy[1], rpy[2]).as_matrix())
T = torch.tensor(pos)
S = torch.tensor(scale)
return OrientedBox(R=R, T=T, S=S)