-
Notifications
You must be signed in to change notification settings - Fork 8
/
main_nuscenes.py
183 lines (152 loc) · 8.26 KB
/
main_nuscenes.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
""" inference on the nuscenes dataset
"""
import os, numpy as np, argparse, json, sys, numba, yaml, multiprocessing, shutil
import mot_3d.visualization as visualization, mot_3d.utils as utils
from mot_3d.data_protos import BBox, Validity
from mot_3d.mot import MOTModel
from mot_3d.frame_data import FrameData
from data_loader import NuScenesLoader
from pyquaternion import Quaternion
from nuscenes.utils.data_classes import Box
parser = argparse.ArgumentParser()
# running configurations
parser.add_argument('--name', type=str, default='immortal')
parser.add_argument('--det_name', type=str, default='cp')
parser.add_argument('--process', type=int, default=1)
parser.add_argument('--visualize', action='store_true', default=False)
parser.add_argument('--skip', action='store_true', default=False)
parser.add_argument('--start_frame', type=int, default=0, help='start at a middle frame for debug')
parser.add_argument('--obj_types', default='car,bus,trailer,truck,pedestrian,bicycle,motorcycle')
# paths
parser.add_argument('--config_path', type=str, default='configs/nu_configs/immortal.yaml')
parser.add_argument('--result_folder', type=str, default='./mot_results/nuscenes')
parser.add_argument('--data_folder', type=str, default='./data/nuscenes/')
parser.add_argument('--det_data_folder', type=str, default='./data/nuscenes/')
parser.add_argument('--test', action='store_true', default=False)
args = parser.parse_args()
def nu_array2mot_bbox(b):
nu_box = Box(b[:3], b[3:6], Quaternion(b[6:10]))
mot_bbox = BBox(
x=nu_box.center[0], y=nu_box.center[1], z=nu_box.center[2],
w=nu_box.wlh[0], l=nu_box.wlh[1], h=nu_box.wlh[2],
o=nu_box.orientation.yaw_pitch_roll[0]
)
if len(b) == 11:
mot_bbox.s = b[-1]
return mot_bbox
def load_gt_bboxes(data_folder, type_token, segment_name):
gt_info = np.load(os.path.join(data_folder, 'gt_info', '{:}.npz'.format(segment_name)), allow_pickle=True)
ids, inst_types, bboxes = gt_info['ids'], gt_info['types'], gt_info['bboxes']
mot_bboxes = list()
for _, frame_bboxes in enumerate(bboxes):
mot_bboxes.append([])
for _, b in enumerate(frame_bboxes):
mot_bboxes[-1].append(BBox.bbox2array(nu_array2mot_bbox(b)))
gt_ids, gt_bboxes = utils.inst_filter(ids, mot_bboxes, inst_types,
type_field=type_token, id_trans=True)
return gt_bboxes, gt_ids
def frame_visualization(bboxes, ids, states, gt_bboxes=None, gt_ids=None, pc=None, dets=None, name=''):
visualizer = visualization.Visualizer2D(name=name, figsize=(12, 12))
if pc is not None:
visualizer.handler_pc(pc)
if gt_bboxes is not None:
for _, bbox in enumerate(gt_bboxes):
visualizer.handler_box(bbox, message='', color='black')
dets = [d for d in dets if d.s >= 0.01]
for det in dets:
visualizer.handler_box(det, message='%.2f' % det.s, color='gray', linestyle='dashed')
for _, (bbox, id, state_string) in enumerate(zip(bboxes, ids, states)):
if Validity.valid(state_string):
visualizer.handler_box(bbox, message='%.2f %s'%(bbox.s, id), color='red')
else:
visualizer.handler_box(bbox, message='%.2f %s'%(bbox.s, id), color='light_blue')
# visualizer.show()
save_path = './{:}.png'.format(name)
visualizer.save(save_path)
visualizer.close()
def sequence_mot(configs, data_loader, obj_type, sequence_id, gt_bboxes=None, gt_ids=None, visualize=False):
tracker = MOTModel(configs)
frame_num = len(data_loader)
IDs, bboxes, states, types = list(), list(), list(), list()
for frame_index in range(data_loader.cur_frame, frame_num):
if frame_index % 10 == 0:
print('TYPE {:} SEQ {:} Frame {:} / {:}'.format(obj_type, sequence_id, frame_index + 1, frame_num))
# input data
frame_data = next(data_loader)
frame_data = FrameData(dets=frame_data['dets'], ego=frame_data['ego'], pc=frame_data['pc'],
det_types=frame_data['det_types'], aux_info=frame_data['aux_info'], time_stamp=frame_data['time_stamp'])
# mot
results = tracker.frame_mot(frame_data)
result_pred_bboxes = [trk[0] for trk in results]
result_pred_ids = [trk[1] for trk in results]
result_pred_states = [trk[2] for trk in results]
result_types = [trk[3] for trk in results]
# visualization
if visualize:
frame_visualization(result_pred_bboxes, result_pred_ids, result_pred_states,
gt_bboxes[frame_index], gt_ids[frame_index], frame_data.pc, dets=frame_data.dets, name='{:}_{:}'.format(args.name, frame_index))
# wrap for output
IDs.append(result_pred_ids)
result_pred_bboxes = [BBox.bbox2array(bbox) for bbox in result_pred_bboxes]
bboxes.append(result_pred_bboxes)
states.append(result_pred_states)
types.append(result_types)
return IDs, bboxes, states, types
def main(name, obj_types, config_path, data_folder, det_data_folder, result_folder, start_frame=0, token=0, process=1):
for obj_type in obj_types:
summary_folder = os.path.join(result_folder, 'summary', obj_type)
# simply knowing about all the segments
file_names = sorted(os.listdir(os.path.join(data_folder, 'ego_info')))
if args.skip:
file_names = [fname for fname in file_names if not os.path.exists(os.path.join(summary_folder, fname))]
# load model configs
configs = yaml.load(open(config_path, 'r'))
for file_index, file_name in enumerate(file_names[:]):
if file_index % process != token:
continue
print('START TYPE {:} SEQ {:} / {:}'.format(obj_type, file_index + 1, len(file_names)))
segment_name = file_name.split('.')[0]
data_loader = NuScenesLoader(configs, [obj_type], segment_name, data_folder, det_data_folder, start_frame)
# if not args.test:
# gt_bboxes, gt_ids = load_gt_bboxes(data_folder, [obj_type], segment_name)
# else:
# gt_bboxes, gt_ids = [None for i in range(len(data_loader))], [None for i in range(len(data_loader))]
ids, bboxes, states, types = sequence_mot(configs, data_loader, obj_type, file_index, None, None, args.visualize)
frame_num = len(ids)
for frame_index in range(frame_num):
id_num = len(ids[frame_index])
for i in range(id_num):
ids[frame_index][i] = '{:}_{:}'.format(file_index, ids[frame_index][i])
np.savez_compressed(os.path.join(summary_folder, '{}.npz'.format(segment_name)),
ids=ids, bboxes=bboxes, states=states, types=types)
if __name__ == '__main__':
if args.test:
args.data_folder = os.path.join(args.data_folder, 'test_2hz')
args.det_data_folder = os.path.join(args.det_data_folder, 'test_2hz', 'detection')
args.result_folder = os.path.join(args.result_folder, 'test_2hz')
else:
args.data_folder = os.path.join(args.data_folder, 'validation_2hz')
args.det_data_folder = os.path.join(args.det_data_folder, 'validation_2hz', 'detection')
args.result_folder = os.path.join(args.result_folder, 'validation_2hz')
result_folder = os.path.join(args.result_folder, args.name)
if not os.path.exists(result_folder):
os.makedirs(result_folder)
summary_folder = os.path.join(result_folder, 'summary')
if not os.path.exists(summary_folder):
os.makedirs(summary_folder)
det_data_folder = os.path.join(args.det_data_folder, args.det_name)
obj_types = args.obj_types.split(',')
for obj_type in obj_types:
tmp_summary_folder = os.path.join(summary_folder, obj_type)
if not os.path.exists(tmp_summary_folder):
os.makedirs(tmp_summary_folder)
if args.process > 1:
pool = multiprocessing.Pool(args.process)
for token in range(args.process):
result = pool.apply_async(main, args=(args.name, obj_types, args.config_path, args.data_folder, det_data_folder,
result_folder, 0, token, args.process))
pool.close()
pool.join()
else:
main(args.name, obj_types, args.config_path, args.data_folder, det_data_folder,
result_folder, args.start_frame, 0, 1)