/
soccernetv3.py
196 lines (155 loc) · 8.53 KB
/
soccernetv3.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
from __future__ import division, print_function, absolute_import
import glob
import os
import os.path as osp
from ..dataset import ImageDataset
from SoccerNet.Downloader import SoccerNetDownloader as SNdl
import zipfile
class Soccernetv3(ImageDataset):
"""Soccernet-v3 train and valid sets. When set as "source" in the run configs (cfg.data.sources), the train set is
used for training. When set as "target" set in the run configs (cfg.data.targets), the valid set is used for performance
evaluation.
"""
dataset_dir = 'soccernetv3'
def __init__(self, root='', soccernetv3_training_subset=1.0, **kwargs):
assert 1.0 >= soccernetv3_training_subset > 0.0
self.root = osp.abspath(osp.expanduser(root))
self.dataset_dir = osp.join(self.root, self.dataset_dir)
self.reid_dataset_dir = self.download_soccernet_dataset(self.dataset_dir, ["valid", "train"])
self.train_dir = osp.join(self.reid_dataset_dir, 'train')
self.query_dir = osp.join(self.reid_dataset_dir, 'valid/query')
self.gallery_dir = osp.join(self.reid_dataset_dir, 'valid/gallery')
required_files = [
self.reid_dataset_dir, self.train_dir, self.query_dir, self.gallery_dir
]
self.check_before_run(required_files)
train, _, _ = self.process_dir(self.train_dir, {}, relabel=True, soccernetv3_training_subset=soccernetv3_training_subset)
query, pid2label, ids_counter = self.process_dir(self.query_dir, {}, 0)
gallery, pid2label, ids_counter = self.process_dir(self.gallery_dir, pid2label, ids_counter)
super(Soccernetv3, self).__init__(train, query, gallery, **kwargs)
def process_dir(self, main_path, pid2label, ids_counter=0, relabel=False, soccernetv3_training_subset=1.):
data = []
img_paths = glob.glob(osp.join(main_path, '*/*/*/*/*.png'))
# sort images list such that each sample position in the list match its filename index
img_paths.sort(key=lambda img_path: self.get_bbox_index(img_path))
# if soccernetv3_training_subset is set, use samples from action '0' to action 'end_action'
action_num = self.extract_sample_info(os.path.basename(img_paths[-1]))["action_idx"] + 1
end_action = action_num * soccernetv3_training_subset
for img_path in img_paths:
filename = os.path.basename(img_path)
info = self.extract_sample_info(filename)
pid = info["person_uid"]
action_idx = info["action_idx"]
if action_idx >= end_action:
break
if relabel:
if pid not in pid2label:
pid2label[pid] = ids_counter
ids_counter += 1
pid = pid2label[pid]
data.append((img_path, pid, action_idx))
return data, pid2label, ids_counter
@staticmethod
def download_soccernet_dataset(dataset_dir, split):
task = "reid"
reid_dataset_dir = osp.join(dataset_dir, task)
mySNdl = SNdl(LocalDirectory=dataset_dir)
for set_type in split:
# download SoccerNet dataset subsets specified by 'set_type' (train/valid/test/challenge)
path_to_set = osp.join(reid_dataset_dir, set_type)
if osp.exists(path_to_set):
print("SoccerNet {} set was already downloaded and unzipped at {}.".format(set_type, path_to_set))
continue
mySNdl.downloadDataTask(task=task, split=[set_type])
print("Unzipping {} set to '{}' ...".format(set_type, reid_dataset_dir))
path_to_zip_file = osp.join(reid_dataset_dir, set_type + ".zip")
if not osp.exists(path_to_zip_file):
raise FileNotFoundError("Missing zip file {}.".format(path_to_zip_file))
else:
with zipfile.ZipFile(path_to_zip_file, "r") as zip_ref:
zip_ref.extractall(reid_dataset_dir)
print("Deleting {} set zip file at '{}'...".format(set_type, path_to_zip_file))
os.remove(path_to_zip_file)
print('SoccerNet {} set is ready.'.format(set_type))
return reid_dataset_dir
@staticmethod
def extract_sample_info(filename):
""" Extract sample annotations from its filename
File naming convention is:
- For public samples (train/valid/test set): '<bbox_idx>-<action_idx>-<person_uid>-<frame_idx>-<clazz>-<id>-<UAI>-<image_size>.png'
- For anonymous samples (challenge set): '<bbox_idx>-<action_idx>-<image_size>.png'
The "id" field is the identifier of the player within an action. When the id is given as a number, it refers
to the player jersey number. The jersey number is provided for a player if it can be seen at least once
within one frame of the action. If the jersey number is not visible in any frame of the action, then this
identifier is given as a letter.
"""
info = {}
splits = filename.split(".")[0].split("-")
if len(splits) == 8:
info["bbox_idx"] = int(splits[0])
info["action_idx"] = int(splits[1])
info["person_uid"] = splits[2]
info["frame_idx"] = int(splits[3])
info["clazz"] = splits[4]
info["id"] = splits[5]
info["UAI"] = splits[6]
shape = splits[7].split("x")
info["shape"] = (int(shape[0]), int(shape[1]))
elif len(splits) == 3:
info["bbox_idx"] = int(splits[0])
info["action_idx"] = int(splits[1])
shape = splits[2].split("x")
info["shape"] = (int(shape[0]), int(shape[1]))
else:
raise ValueError("Wrong sample filename format '{}'".format(filename))
return info
@staticmethod
def get_bbox_index(filepath):
return int(os.path.basename(filepath).split("-")[0])
class Soccernetv3Test(Soccernetv3):
""" Soccernet-v3 test set. Can be used as "target" set in the run configs (cfg.data.targets) for performance evaluation.
"""
def __init__(self, root='', **kwargs):
self.root = osp.abspath(osp.expanduser(root))
self.dataset_dir = osp.join(self.root, self.dataset_dir)
self.reid_dataset_dir = self.download_soccernet_dataset(self.dataset_dir, ["test"])
self.query_dir = osp.join(self.reid_dataset_dir, 'test/query')
self.gallery_dir = osp.join(self.reid_dataset_dir, 'test/gallery')
required_files = [
self.reid_dataset_dir, self.query_dir, self.gallery_dir
]
self.check_before_run(required_files)
train = []
query, pid2label, ids_counter = self.process_dir(self.query_dir, {}, 0)
gallery, _, _ = self.process_dir(self.gallery_dir, pid2label, ids_counter)
super(Soccernetv3, self).__init__(train, query, gallery, **kwargs)
class Soccernetv3Challenge(Soccernetv3):
""" Soccernet-v3 challenge set. Can be used as "target" set in the run configs (cfg.data.targets) together with the
export_ranking_results config (cfg.test.export_ranking_results) in order to export ranking results as a JSON file
for external evaluation.
"""
hidden_labels = True
def __init__(self, root='', **kwargs):
self.root = osp.abspath(osp.expanduser(root))
self.dataset_dir = osp.join(self.root, self.dataset_dir)
self.reid_dataset_dir = self.download_soccernet_dataset(self.dataset_dir, ["challenge"])
self.query_dir = osp.join(self.reid_dataset_dir, 'challenge/query')
self.gallery_dir = osp.join(self.reid_dataset_dir, 'challenge/gallery')
required_files = [
self.reid_dataset_dir, self.query_dir, self.gallery_dir
]
self.check_before_run(required_files)
train = []
query, pid2label, ids_counter = self.process_dir(self.query_dir, {}, 0)
gallery, _, _ = self.process_dir(self.gallery_dir, pid2label, ids_counter)
super(Soccernetv3, self).__init__(train, query, gallery, **kwargs)
def process_dir(self, main_path, pid2label, ids_counter=0, relabel=False, soccernetv3_training_subset=1.):
data = []
img_paths = glob.glob(osp.join(main_path, '*.png'))
# sort images list such that each sample position in the list match its filename index
img_paths.sort(key=lambda img_path: self.get_bbox_index(img_path))
for img_path in img_paths:
filename = os.path.basename(img_path)
info = self.extract_sample_info(filename)
data.append((img_path, (info["bbox_idx"]), info["action_idx"]))
return data, pid2label, ids_counter