/
track2reid.py
196 lines (152 loc) · 7.12 KB
/
track2reid.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
"""
@author: tenghehan
处理 tracking 结果，生成对应的 reid 数据集.
将 id 按照 partition rate 随机划分成 train 和 test.
"""
import os
import shutil
import cv2
import argparse
import random
from tqdm import tqdm
from utils.log import get_logger
from utils.txt_logger import txt_logger
class Range(object):
def __init__(self, start, end):
self.start = start
self.end = end
def __eq__(self, other):
return self.start <= other <= self.end
class ReIDDataConverter():
def __init__(self, args):
self.logger = get_logger('root')
self.txt_logger = txt_logger(os.path.join(args.save_path, args.dataset_name, 'info.txt'))
self.dataset_name = args.dataset_name
self.images_path = os.path.join(args.image_sequence_path, self.dataset_name, "img1")
assert os.path.isdir(self.images_path), "Images path error"
self.track_result_path = os.path.join(args.track_result_path, f'{self.dataset_name}.txt')
assert os.path.isfile(self.track_result_path), "Tracking result path error"
self.save_path = os.path.join(args.save_path, self.dataset_name)
if os.path.exists(self.save_path):
shutil.rmtree(self.save_path)
os.makedirs(os.path.join(self.save_path, 'train'), exist_ok=True)
os.makedirs(os.path.join(self.save_path, 'test'), exist_ok=True)
os.makedirs(os.path.join(self.save_path, 'query'), exist_ok=True)
self.sampling_rate = args.sampling_rate
self.sampled_imgs_filenames = self.sample_frames()
self.partition_rate = args.partition_rate
self.track_result = []
self.id_set = {'train_id_set': set(), 'test_id_set': set()}
self.id_list = []
self.id_images_details = {}
def partition_train_test(self, id_list):
total_num = len(id_list)
train_list = random.sample(id_list, int(total_num * self.partition_rate))
train_set = set(train_list)
self.id_set = {
'train_id_set': train_set,
'test_id_set': set(id_list) - train_set,
}
for identity in self.id_list:
self.id_images_details[identity] = []
def sample_frames(self):
imgs_filenames = sorted(os.listdir(self.images_path))
sampled_imgs_filenames = []
for img_filename in imgs_filenames:
if random.random() <= self.sampling_rate:
sampled_imgs_filenames.append(img_filename)
return sampled_imgs_filenames
def process_track_result(self):
for line in open(self.track_result_path):
info = line.split(',')
idx_frame = int(info[0])
identity = int(info[1])
# bbox: tlwh
bbox = (int(info[2]), int(info[3]), int(info[4]), int(info[5]))
info_dict = {
'idx_frame': idx_frame,
'identity': identity,
'bbox': bbox
}
self.track_result.append(info_dict)
if identity not in self.id_list:
self.id_list.append(identity)
self.partition_train_test(self.id_list)
return self.track_result, self.id_set
def cal_train_test_ids(self):
train_ids, test_ids = 0, 0
for id in self.id_set['train_id_set']:
if len(self.id_images_details[id]) > 0:
train_ids += 1
for id in self.id_set['test_id_set']:
if len(self.id_images_details[id]) > 0:
test_ids += 1
return train_ids, test_ids
def generate_reid_dataset(self):
trainset_size = 0
testset_size = 0
for track in tqdm(self.track_result):
# if the frame is sampled
idx_frame = track['idx_frame']
if f'{str(idx_frame).zfill(6)}.jpg' not in self.sampled_imgs_filenames:
continue
# read frame image
frame_path = os.path.join(self.images_path, f'{str(idx_frame).zfill(6)}.jpg')
frame = cv2.imread(frame_path)
# crop the person area from the whole image
x1 = track['bbox'][0]
y1 = track['bbox'][1]
x2 = track['bbox'][0] + track['bbox'][2]
y2 = track['bbox'][1] + track['bbox'][3]
if x2 <= x1 or y2 <= y1:
continue
cropped_img = frame[y1:y2, x1:x2]
img_name = str(track['identity']).zfill(5) + '_c1_' + str(idx_frame).zfill(6) + '.jpg'
self.id_images_details[track['identity']].append(img_name)
# save the person image into reid train/test dataset
if track['identity'] in self.id_set['train_id_set']:
cv2.imwrite(os.path.join(self.save_path, 'train', img_name), cropped_img)
trainset_size += 1
else:
cv2.imwrite(os.path.join(self.save_path, 'test', img_name), cropped_img)
testset_size += 1
train_ids, test_ids = self.cal_train_test_ids()
return train_ids, test_ids, trainset_size, testset_size
def select_query_images(self):
query_size = 0
for identity in self.id_set['test_id_set']:
if len(self.id_images_details[identity]) > 1:
shutil.move(os.path.join(self.save_path, 'test', self.id_images_details[identity][0]),
os.path.join(self.save_path, 'query', self.id_images_details[identity][0].replace("_c1_", "_c2_")))
query_size += 1
return query_size
def run(self):
self.txt_logger.add_info('sampling rate: {}'.format(self.sampling_rate))
self.txt_logger.add_info('partition rate: {}'.format(self.partition_rate))
self.track_result, self.id_set = self.process_track_result()
self.logger.info('generating reid dataset ...')
train_ids, test_ids, trainset_size, testset_size = self.generate_reid_dataset()
self.logger.info('selecting query images from test dataset ...')
query_size = self.select_query_images()
self.logger.info('reid dataset {} generated'.format(self.dataset_name))
self.txt_logger.add_info('trainset: {} identities, {} images'.format(
train_ids, trainset_size))
self.txt_logger.add_info('testset: {} identities, {} images'.format(
test_ids, (testset_size - query_size)))
self.txt_logger.add_info('query: {} images'.format(query_size))
self.txt_logger.output()
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument("--image_sequence_path", type=str, default="./image_sequence/")
parser.add_argument("--track_result_path", type=str, default="./output/")
parser.add_argument("--save_path", type=str, default="./reid_dataset/")
parser.add_argument("--sampling_rate", type=float, default=1, choices=[Range(0.0, 1.0)])
parser.add_argument("--dataset_name", type=str)
parser.add_argument("--partition_rate", type=float, default=0.8, choices=[Range(0.0, 1.0)], help="percentage (identity) of training set")
return parser.parse_args()
if __name__ == "__main__":
rand_seed = 50
random.seed(rand_seed)
args = parse_args()
reidDataConverter = ReIDDataConverter(args)
reidDataConverter.run()