forked from mit-han-lab/pvcnn
-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
140 lines (123 loc) · 5.05 KB
/
utils.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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
# Copyright 2018 Charles R. Qi from Stanford University and
# Wei Liu from Nuro Inc.
#
# 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.
import numba
import numpy as np
from scipy.spatial import ConvexHull
__all__ = ['get_box_iou_3d']
@numba.njit()
def poly_area(coords):
"""
calculate area of polygon given x-y coordinates
(ref: http://stackoverflow.com/questions/24467972/calculate-area-of-polygon-given-x-y-coordinates)
:param coords: FloatTensor[4, 2]
"""
x = coords[:, 0]
y = coords[:, 1]
return 0.5 * np.abs(np.dot(x, np.roll(y, 1)) - np.dot(y, np.roll(x, 1)))
def polygon_clip(subject_polygon, clip_polygon):
"""
clip a polygon with another polygon
(ref: https://rosettacode.org/wiki/Sutherland-Hodgman_polygon_clipping#Python)
:param subject_polygon: a list of (x,y) 2d points, any polygon
:param clip_polygon: a list of (x,y) 2d points, has to be *convex*
:return:
a list of (x,y) vertex point for the intersection polygon
"""
def inside(p):
return (cp2[0] - cp1[0]) * (p[1] - cp1[1]) > (cp2[1] - cp1[1]) * (p[0] - cp1[0])
def compute_intersection():
dc = [cp1[0] - cp2[0], cp1[1] - cp2[1]]
dp = [s[0] - e[0], s[1] - e[1]]
n1 = cp1[0] * cp2[1] - cp1[1] * cp2[0]
n2 = s[0] * e[1] - s[1] * e[0]
n3 = 1.0 / (dc[0] * dp[1] - dc[1] * dp[0])
return [(n1 * dp[0] - n2 * dc[0]) * n3, (n1 * dp[1] - n2 * dc[1]) * n3]
output_list = subject_polygon
cp1 = clip_polygon[-1]
for clip_vertex in clip_polygon:
cp2 = clip_vertex
input_list = output_list
output_list = []
s = input_list[-1]
for subject_vertex in input_list:
e = subject_vertex
if inside(e):
if not inside(s):
output_list.append(compute_intersection())
output_list.append(e)
elif inside(s):
output_list.append(compute_intersection())
s = e
cp1 = cp2
if len(output_list) == 0:
return None
return output_list
def convex_hull_intersection(p1, pt):
"""
compute area of two convex hull's intersection area
:param p1: a list of (x,y) tuples of hull vertices
:param pt: a list of (x,y) tuples of hull vertices
:return:
a list of (x,y) for the intersection and its volume
"""
inter_p = polygon_clip(p1, pt)
if inter_p is not None:
hull_inter = ConvexHull(inter_p)
return inter_p, hull_inter.volume
else:
return None, 0.0
@numba.njit()
def box_volume_3d(corners):
a = np.sqrt(np.sum((corners[:, 0] - corners[:, 1]) ** 2))
b = np.sqrt(np.sum((corners[:, 1] - corners[:, 2]) ** 2))
c = np.sqrt(np.sum((corners[:, 0] - corners[:, 4]) ** 2))
return a * b * c
def get_box_iou_3d(corners_1, corners_t):
"""
calculate iou of 3d box
:param corners_1: FloatTensor[B, 3, 8], assume up direction is negative Y
:param corners_t: FloatTensor[B, 3, 8], assume up direction is negative Y
NOTE: corner points are in counter clockwise order, e.g.,
2--1
3--0 5
7--4
:return:
iou_3d: 3D bounding box IoU, FloatTensor[B]
iou_2d: bird's eye view 2D bounding box IoU, FloatTensor[B]
"""
if corners_1.ndim == 3:
batch_size = corners_1.shape[0]
iou_3d = np.zeros(batch_size)
iou_2d = np.zeros(batch_size)
for b in range(batch_size):
iou_3d[b], iou_2d[b] = get_box_iou_3d(corners_1[b], corners_t[b])
return iou_3d, iou_2d
else:
# corner points are in counter clockwise order
corners_1_upper_xz = [(corners_1[0, 3], corners_1[2, 3]), (corners_1[0, 2], corners_1[2, 2]),
(corners_1[0, 1], corners_1[2, 1]), (corners_1[0, 0], corners_1[2, 0])]
corners_t_upper_xz = [(corners_t[0, 3], corners_t[2, 3]), (corners_t[0, 2], corners_t[2, 2]),
(corners_t[0, 1], corners_t[2, 1]), (corners_t[0, 0], corners_t[2, 0])]
area_1 = poly_area(np.array(corners_1_upper_xz))
area_2 = poly_area(np.array(corners_t_upper_xz))
inter, inter_area = convex_hull_intersection(corners_1_upper_xz, corners_t_upper_xz)
iou_2d = inter_area / (area_1 + area_2 - inter_area)
y_max = min(corners_1[1, 0], corners_t[1, 0])
y_min = max(corners_1[1, 4], corners_t[1, 4])
inter_vol = inter_area * max(0.0, y_max - y_min)
vol1 = box_volume_3d(corners_1)
vol2 = box_volume_3d(corners_t)
iou_3d = inter_vol / (vol1 + vol2 - inter_vol)
return iou_3d, iou_2d