-
Notifications
You must be signed in to change notification settings - Fork 17
/
create_range_image_in_kitti.py
293 lines (250 loc) · 11 KB
/
create_range_image_in_kitti.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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
#keep full resolution range image
import numpy as np
import pickle as pkl
from pdb import set_trace
import os
from kitti_utils import Calibration
from pathlib import Path
import matplotlib.pyplot as plt
import tqdm
import collections
import argparse
from queue import Queue
from threading import Thread
def parse_args():
parser = argparse.ArgumentParser(description='Create range images in KITTI')
parser.add_argument('--source-dir', help='path to KITTI in MMDet3D format', type=str)
parser.add_argument('--target-dir', help='path to save the extracted data')
parser.add_argument('--num-threads', help='path to save the extracted data', type=int, default=10)
args = parser.parse_args()
return args
def boxes3d_kitti_camera_to_lidar(boxes3d_camera, calib):
"""
Args:
boxes3d_camera: (N, 7) [x, y, z, l, h, w, r] in rect camera coords
calib:
Returns:
boxes3d_lidar: [x, y, z, dx, dy, dz, heading], (x, y, z) is the box center
"""
xyz_camera = boxes3d_camera[:, 0:3]
l, h, w, r = boxes3d_camera[:, 3:4], boxes3d_camera[:, 4:5], boxes3d_camera[:, 5:6], boxes3d_camera[:, 6:7]
xyz_lidar = calib.rect_to_lidar(xyz_camera)
xyz_lidar[:, 2] += h[:, 0] / 2
return np.concatenate([xyz_lidar, l, w, h, -(r + np.pi / 2)], axis=-1)
def to_xyz0z1(bbox_type7): #[n, r, 7]
batch_size, num_bbox, _ = bbox_type7.shape
dtype = bbox_type7.dtype
xy_4pts = np.full((batch_size, num_bbox, 4, 2), 0, dtype = dtype)
xy_4pts[:,:,0,:] = np.array([[[ 0.5, -0.5]]], dtype = dtype) * bbox_type7[:,:,3:5]
xy_4pts[:,:,1,:] = np.array([[[-0.5, -0.5]]], dtype = dtype) * bbox_type7[:,:,3:5]
xy_4pts[:,:,2,:] = np.array([[[-0.5, 0.5]]], dtype = dtype) * bbox_type7[:,:,3:5]
xy_4pts[:,:,3,:] = np.array([[[ 0.5, 0.5]]], dtype = dtype) * bbox_type7[:,:,3:5]
cosa = np.cos(bbox_type7[:,:,-1])
sina = np.sin(bbox_type7[:,:,-1])
rot_mat = np.stack([cosa, -sina, sina, cosa], axis = -1).reshape(batch_size, num_bbox, 2, 2)
rot_4pts = np.einsum('nrij,nrmj->nrmi', rot_mat, xy_4pts)
rot_4pts = rot_4pts + bbox_type7[:,:,None,:2]
rot_4pts = rot_4pts.reshape(batch_size, num_bbox, 8)
z0 = bbox_type7[:,:,2] - bbox_type7[:,:,5] / 2
z1 = bbox_type7[:,:,2] + bbox_type7[:,:,5] / 2
bbox_xyz0z1 = np.concatenate([rot_4pts, z0[:,:,None], z1[:,:,None]], axis = 2)
return bbox_xyz0z1
def to_8pts(bbox_4pts):
bbox_4pts = bbox_4pts.astype(np.float32)
xy = bbox_4pts[:,:8].reshape(-1,4,2)
z_bot = bbox_4pts[:,8]
z_bot = np.tile(z_bot[:,None],(1,4))
z_top = bbox_4pts[:,9]
z_top = np.tile(z_top[:,None],(1,4))
xyz_bot = np.concatenate([xy, z_bot[:,:,None]],axis = 2)
xyz_top = np.concatenate([xy, z_top[:,:,None]],axis = 2)
bbox_8pts = np.concatenate([xyz_bot,xyz_top], axis = 1)
return bbox_8pts
def name_to_cls(names):
cls_mapping = {'Car':1, 'Pedestrian':2, 'Cyclist':4}
gt_class = []
for name in names:
if name in cls_mapping:
gt_class.append(cls_mapping[name])
else:
gt_class.append(-1)
gt_class = np.array(gt_class)
return gt_class
def get_pc(target_dir, pc_idx, is_test=False):
if is_test:
path = '{}/testing/velodyne/{}.bin'.format(target_dir, pc_idx)
else:
path = '{}/training/velodyne/{}.bin'.format(target_dir, pc_idx)
pc = np.fromfile(path, dtype = np.float32).reshape(-1, 4)
return pc
def get_gt_bbox(location, dimensions, rotation_y, calib):
'''
location: object location x,y,z in camera coordinates (in meters)
dimensions 3D object dimensions: height, width, length (in meters)
rotation_y Rotation ry around Y-axis in camera coordinates [-pi..pi]
'''
gt_bbox_camera = np.concatenate([location, dimensions, rotation_y[:,None]], axis = 1).astype(np.float32)
gt_bbox_lidar = boxes3d_kitti_camera_to_lidar(gt_bbox_camera, calib)
bbox_xyz0z1 = to_xyz0z1(gt_bbox_lidar[None, :, :]).squeeze(0)
bbox_8pts = to_8pts(bbox_xyz0z1)
return bbox_8pts
def get_range_image(pc, incl, height):
incl_deg = incl * 180 / 3.1415
# print(incl - np.roll(incl, 1))
xy_norm = np.linalg.norm(pc[:, :2], ord = 2, axis = 1)
error_list = []
for i in range(len(incl)):
h = height[i]
theta = incl[i]
error = np.abs(theta - np.arctan2(h - pc[:,2], xy_norm))
error_list.append(error)
all_error = np.stack(error_list, axis=-1)
row_inds = np.argmin(all_error, axis=-1)
azi = np.arctan2(pc[:,1], pc[:,0])
width = 2048
col_inds = width - 1.0 + 0.5 - (azi + np.pi) / (2.0 * np.pi) * width
col_inds = np.round(col_inds).astype(np.int32)
col_inds[col_inds == width] = width - 1
col_inds[col_inds < 0] = 0
empty_range_image = np.full((64, width, 5), -1, dtype = np.float32)
point_range = np.linalg.norm(pc[:,:3], axis = 1, ord = 2)
order = np.argsort(-point_range)
point_range = point_range[order]
pc = pc[order]
row_inds = row_inds[order]
col_inds = col_inds[order]
empty_range_image[row_inds, col_inds, :] = np.concatenate([point_range[:,None], pc], axis = 1)
return empty_range_image
def get_calib(source_dir, idx, is_test):
# if is_test:
# path = '/mnt/truenas/scratch/zhichao.li/Data/KITTI/testing/calib'
# else:
# path = '/mnt/truenas/scratch/zhichao.li/Data/KITTI/training/calib'
if is_test:
path = os.path.join(source_dir, 'testing/calib')
else:
path = os.path.join(source_dir, 'training/calib')
calib_file = os.path.join(path, '{}.txt'.format(idx))
p = Path(calib_file)
assert p.exists()
return Calibration(p)
def crop_range_image(range_image):
# width = 2083 // 4
mid = 2083 // 2
beg = mid - 256
end = mid + 256
return range_image[:,beg:end,:]
def process_single_frame(frame, source_dir, target_dir, split, roidb_list):
pc_idx = frame['point_cloud']['lidar_idx']
if split != 'test':
calib = get_calib(source_dir, pc_idx, split=='test')
annos = frame['annos']
gt_class = name_to_cls(annos['name'])
gt_bbox = get_gt_bbox(annos['location'], annos['dimensions'], annos['rotation_y'], calib)
else:
gt_class = np.ones(0,dtype=np.float32)
gt_bbox = np.zeros(0,dtype=np.float32)
pc = get_pc(source_dir, pc_idx, split=='test')
pc_url = os.path.join(target_dir, '{}/{}.npz'.format(npz_dirname, pc_idx))
range_image = get_range_image(pc, incl, height)
range_image_mask = range_image[..., 0] > -1
roidb = {
'gt_class':gt_class,
'gt_bbox_imu':gt_bbox,
'pc_url':pc_url
}
roidb_list.append(roidb)
np.savez(
pc_url,
range_image=range_image,
range_image_mask=range_image_mask,
)
def process_task_worker(frame_queue, source_dir, target_dir, split, roidb_list):
while True:
qsize = frame_queue.qsize()
if qsize > 0:
if qsize % 10 == 0:
print('{} {} frames left'.format(qsize, split))
frame = frame_queue.get()
else:
print("No task left, break down.")
break
try:
process_single_frame(frame, source_dir, target_dir, split, roidb_list)
except Exception as e:
print('Error: ', e)
continue
if __name__ == '__main__':
# KITTI scanning parameters, obtained from Hough transformation
height = np.array(
[0.20966667, 0.2092 , 0.2078 , 0.2078 , 0.2078 ,
0.20733333, 0.20593333, 0.20546667, 0.20593333, 0.20546667,
0.20453333, 0.205 , 0.2036 , 0.20406667, 0.2036 ,
0.20313333, 0.20266667, 0.20266667, 0.20173333, 0.2008 ,
0.2008 , 0.2008 , 0.20033333, 0.1994 , 0.20033333,
0.19986667, 0.1994 , 0.1994 , 0.19893333, 0.19846667,
0.19846667, 0.19846667, 0.12566667, 0.1252 , 0.1252 ,
0.12473333, 0.12473333, 0.1238 , 0.12333333, 0.1238 ,
0.12286667, 0.1224 , 0.12286667, 0.12146667, 0.12146667,
0.121 , 0.12053333, 0.12053333, 0.12053333, 0.12006667,
0.12006667, 0.1196 , 0.11913333, 0.11866667, 0.1182 ,
0.1182 , 0.1182 , 0.11773333, 0.11726667, 0.11726667,
0.1168 , 0.11633333, 0.11633333, 0.1154 ])
zenith = np.array([
0.03373091, 0.02740409, 0.02276443, 0.01517224, 0.01004049,
0.00308099, -0.00155868, -0.00788549, -0.01407172, -0.02103122,
-0.02609267, -0.032068 , -0.03853542, -0.04451074, -0.05020488,
-0.0565317 , -0.06180405, -0.06876355, -0.07361411, -0.08008152,
-0.08577566, -0.09168069, -0.09793721, -0.10398284, -0.11052055,
-0.11656618, -0.12219002, -0.12725147, -0.13407038, -0.14067839,
-0.14510716, -0.15213696, -0.1575499 , -0.16711043, -0.17568678,
-0.18278688, -0.19129293, -0.20247031, -0.21146846, -0.21934183,
-0.22763699, -0.23536977, -0.24528179, -0.25477201, -0.26510582,
-0.27326038, -0.28232882, -0.28893683, -0.30004392, -0.30953414,
-0.31993824, -0.32816311, -0.33723155, -0.34447224, -0.352908 ,
-0.36282001, -0.37216965, -0.38292524, -0.39164219, -0.39895318,
-0.40703745, -0.41835542, -0.42777535, -0.43621111
])
incl = -zenith
args = parse_args()
data_splits = ['training', 'validation', 'test']
source_dir = os.path.abspath(args.source_dir)
target_dir = os.path.abspath(args.target_dir)
os.makedirs(target)
num_threads = args.num_threads
for split in data_splits:
if split == 'training':
npz_dirname = 'npz_trainval'
info_path = os.path.join(source_dir, 'kitti_infos_train.pkl')
elif split == 'validation':
npz_dirname = 'npz_trainval'
info_path = os.path.join(source_dir, 'kitti_infos_val.pkl')
elif split == 'test':
npz_dirname = 'npz_test'
info_path = os.path.join(source_dir, 'kitti_infos_test.pkl')
npz_dirpath = os.path.join(target_dir, npz_dirname)
os.makedirs(npz_dirpath)
# os.makedirs(npz_dirpath, exist_ok=True)
print(f'Begin processing {split} split, and all created data will be saved under: {target_dir}')
data_set = pkl.load(open(info_path, 'rb'))
roidb_list = []
frame_queue = Queue()
for i, frame in enumerate(data_set):
frame_queue.put(frame)
workers = [
Thread(target=process_task_worker, args=(frame_queue, source_dir, target_dir, split, roidb_list))
for _ in range(num_threads)]
for w in workers:
w.start()
for w in workers:
w.join()
print(f'Got {len(roidb_list)} frame in {split} split.')
if split == 'training':
with open(os.path.join(target_dir, 'training.roidb'), 'wb') as fw:
pkl.dump(roidb_list, fw)
elif split == 'validation':
with open(os.path.join(target_dir, 'validation.roidb'), 'wb') as fw:
pkl.dump(roidb_list, fw)
elif split == 'test':
with open(os.path.join(target_dir, 'test.roidb'), 'wb') as fw:
pkl.dump(roidb_list, fw)