-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathmain_brad_test.py
320 lines (283 loc) · 12.6 KB
/
main_brad_test.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
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
import pickle
import torch
import torch.nn as nn
import torch.backends.cudnn as cudnn
import warnings
import random
import numpy as np
import time
import os
from utils.helpers import getTransforms
from main_brad import init_ddp
import moco.builder as mb
from config import parser, setup
from utils import domainnet
import torch.nn.functional as func
import torch.utils.data
from tqdm import tqdm
from sklearn.linear_model import SGDClassifier, LogisticRegression
from sklearn.neighbors import NearestNeighbors
import torch.distributed as dist
def get_argparse():
parser.add_argument('--src-domain', type=str, help='path to file with source domain image paths and classes')
parser.add_argument('--dst-domain', type=str, help='path to file with target domain image paths and classes')
parser.add_argument('--src_save_pth', type=str, default='',
help='path for saving source features. If empty nothing will be saved')
parser.add_argument('--dst_save_pth', type=str, default='',
help='path for saving target features. If empty nothing will be saved')
parser.add_argument('--classifier', default='retrieval', choices=["retrieval", "sgd", "logistic"], help='{"retrieval", "sgd", "logistic"}')
return parser
class AverageMeter(object):
"""Computes and stores the average and current value"""
def __init__(self, name, fmt=':f'):
self.name = name
self.fmt = fmt
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
self.avg = self.sum / self.count
def __str__(self):
fmtstr = '{name} {val' + self.fmt + '} ({avg' + self.fmt + '})'
return fmtstr.format(**self.__dict__)
class ProgressMeter(object):
def __init__(self, num_batches, meters, prefix=""):
self.batch_fmtstr = self._get_batch_fmtstr(num_batches)
self.meters = meters
self.prefix = prefix
def display(self, batch):
entries = [self.prefix + self.batch_fmtstr.format(batch)]
entries += [str(meter) for meter in self.meters]
print('\t'.join(entries))
@staticmethod
def _get_batch_fmtstr(num_batches):
num_digits = len(str(num_batches // 1))
fmt = '{:' + str(num_digits) + 'd}'
return '[' + fmt + '/' + fmt.format(num_batches) + ']'
def cls_metrics(test_features, test_labels, cls):
acc1 = cls.score(test_features, test_labels)
return acc1, 0, 0, 0, 0, 0, 0, 0
def ret_metrics(test_features, test_labels, train_labels, nearest_neigh):
_, top_n_matches_ids = nearest_neigh.kneighbors(test_features)
top_n_labels = train_labels[top_n_matches_ids]
correct = test_labels[:, None] == top_n_labels
acc1 = correct[:, 0:1].any(-1).mean()
acc10 = correct[:, 0:10].any(-1).mean()
acc20 = correct[:, 0:20].any(-1).mean()
p1 = correct[:, 0:1].sum(-1).mean()
p10 = correct[:, 0:10].sum(-1).mean() / 10
p20 = correct[:, 0:20].sum(-1).mean() / 20
p5 = correct[:, 0:5].sum(-1).mean() / 5
p15 = correct[:, 0:15].sum(-1).mean() / 15
return acc1, acc10, acc20, p1, p5, p10, p15, p20
def main(args):
init_ddp(args.local_rank)
args.gpu = args.local_rank
print(dist.get_rank())
is_root = dist.get_rank() == 0
torch.cuda.set_device(args.local_rank)
# setup stuff
args = setup(args)
# seeding the random if required
if args.seed is not None:
random.seed(args.seed)
torch.manual_seed(args.seed)
cudnn.deterministic = True
warnings.warn('You have chosen to fix the CUDNN random seed. '
'This will turn on the CUDNN deterministic setting, '
'which can slow you down considerably! '
'You may see unexpected behavior when restarting '
'from checkpoints.')
# setup transforms
trans, _ = getTransforms(args.edges_sigma, args.process_scale, args.no_cc, args.img2_is_sketch, args.edges_type)
# loading model
model, cur_epoch = mb.load_model(args, return_epoch=args.resume)
device = next(model.parameters()).device
model.isedge = False
features_dim = 2048
# loading source domain data
print(f"Source domain {args.src_domain}")
if os.path.exists(args.src_save_pth):
with open(args.src_save_pth, 'rb') as fp:
img_paths, train_features, train_labels = pickle.load(fp)
else:
file_list_path = args.src_domain
train_dataset = domainnet.Dataset(
file_list_path,
root=os.path.split(file_list_path)[0],
test=True,
transform=trans,
n_samples_per_class=-1
)
ddp_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset, shuffle=False)
train_loader = torch.utils.data.DataLoader(
train_dataset, batch_size=args.batch_size, shuffle=False,
sampler=ddp_sampler,
num_workers=args.workers, pin_memory=True, drop_last=False)
num_train_samples = len(train_dataset)
train_features = np.zeros((num_train_samples, features_dim), dtype=np.float16)
train_labels = np.zeros(num_train_samples, dtype=np.int64)
img_paths = []
end_ind = 0
for ind, (x, y, img_path) in tqdm(enumerate(train_loader)):
with torch.no_grad():
_, _, features = model(x.to(device))
features = torch.mean(features, dim=[-2, -1])
features = nn.functional.normalize(features, dim=1)
features = mb.concat_all_gather_interlace(features)
y = mb.concat_all_gather_interlace(y.to(device))
img_path = mb.concat_all_gather_object_interlace(img_path)
img_paths.extend(img_path)
if not is_root:
continue
y = y.cpu().numpy().astype(np.float16)
features = features.cpu().numpy().astype(np.float16)
begin_ind = end_ind
end_ind = min(begin_ind + len(features), len(train_features))
train_features[begin_ind:end_ind, :] = features[:end_ind - begin_ind]
train_labels[begin_ind:end_ind] = y[:end_ind - begin_ind]
if end_ind >= num_train_samples:
break
if is_root:
if args.src_save_pth:
if os.path.split(args.src_save_pth)[0]:
os.makedirs(os.path.split(args.src_save_pth)[0], exist_ok=True)
with open(args.src_save_pth, 'wb') as fp:
pickle.dump((img_paths, train_features, train_labels), fp, pickle.HIGHEST_PROTOCOL)
if is_root:
if args.classifier == 'sgd':
cls = SGDClassifier(max_iter=1000, n_jobs=16, tol=1e-3).fit(train_features, train_labels)
elif args.classifier == 'logistic':
cls = LogisticRegression(max_iter=1000, n_jobs=16, tol=1e-3).fit(train_features, train_labels)
elif args.classifier == 'retrieval':
cls = NearestNeighbors(n_neighbors=min(20, train_features.shape[0]), algorithm='auto',
n_jobs=-1, metric='correlation').fit(train_features)
else:
raise NotImplementedError()
else:
cls = None
print(f"Target domain {args.dst_domain}")
batch_time = AverageMeter('Time', ':6.5f')
acc1 = AverageMeter('Acc@1', ':6.5f')
acc10 = AverageMeter('Acc@10', ':6.5f')
acc20 = AverageMeter('Acc@20', ':6.5f')
precision1 = AverageMeter('p@1', ':6.5f')
precision10 = AverageMeter('p@10', ':6.5f')
precision20 = AverageMeter('p@20', ':6.5f')
precision5 = AverageMeter('p@5', ':6.3f')
precision15 = AverageMeter('p@15', ':6.3f')
if os.path.exists(args.dst_save_pth):
with open(args.dst_save_pth, 'rb') as fp:
saved_features = pickle.load(fp)
if args.is_root:
if args.classifier == 'retrieval':
a1, a10, a20, p1, p5, p10, p15, p20 = ret_metrics(saved_features[1], saved_features[2],
train_labels, cls)
else:
a1, a10, a20, p1, p5, p10, p15, p20 = cls_metrics(saved_features[1], saved_features[2],
cls)
else:
a1, a10, a20, p1, p5, p10, p15, p20 = 0., 0., 0., 0., 0., 0., 0., 0.
y = saved_features[2]
acc1.update(a1, len(y))
acc10.update(a10, len(y))
acc20.update(a20, len(y))
precision1.update(p1, len(y))
precision5.update(p5, len(y))
precision10.update(p10, len(y))
precision15.update(p15, len(y))
precision20.update(p20, len(y))
else:
file_list_path = args.dst_domain
test_dataset = domainnet.Dataset(
file_list_path,
test=True,
root=os.path.split(file_list_path)[0],
transform=trans,
)
test_sampler = torch.utils.data.distributed.DistributedSampler(test_dataset, shuffle=False)
test_loader = torch.utils.data.DataLoader(
test_dataset, batch_size=args.batch_size, shuffle=False, sampler=test_sampler,
num_workers=args.workers, pin_memory=True)
progress = ProgressMeter(
len(test_loader),
[batch_time, acc1, acc10, acc20, precision1, precision5, precision10, precision15, precision20],
prefix=f"Train on {os.path.basename(args.src_domain)} Test on {os.path.basename(args.dst_domain)}")
end = time.time()
num_test_samples = len(test_dataset)
print(f'num_test_samples: {num_test_samples}')
total_samples = 0
img_paths = []
all_features = []
all_y = []
for ind, (x, y, img_path) in enumerate(test_loader):
with torch.no_grad():
_, _, features = model(x.to(device))
features = torch.mean(features, dim=[-2, -1])
features = nn.functional.normalize(features, dim=1)
features = mb.concat_all_gather_interlace(features)
y = mb.concat_all_gather_interlace(y.to(device)).cpu()
img_path = mb.concat_all_gather_object_interlace(img_path)
img_paths.extend(img_path)
if not is_root:
continue
features = features.cpu().numpy().astype(np.float16)
y = y.numpy()
total_samples += len(y)
if total_samples > num_test_samples:
diff = total_samples - num_test_samples
features = features[:-diff]
y = y[:-diff]
all_features.append(features)
all_y.append(y)
print(f'diff {diff}')
all_features.append(features)
all_y.append(y)
if args.classifier == 'retrieval':
a1, a10, a20, p1, p5, p10, p15, p20 = ret_metrics(features, y, train_labels, cls)
else:
a1, a10, a20, p1, p5, p10, p15, p20 = cls_metrics(features, y, cls)
acc1.update(a1, len(y))
acc10.update(a10, len(y))
acc20.update(a20, len(y))
precision1.update(p1, len(y))
precision10.update(p10, len(y))
precision20.update(p20, len(y))
precision5.update(p5, len(y))
precision15.update(p15, len(y))
batch_time.update(time.time() - end)
end = time.time()
if ind % 10 == 0:
progress.display(ind)
if args.dst_save_pth:
if os.path.split(args.dst_save_pth)[0]:
os.makedirs(os.path.split(args.dst_save_pth)[0], exist_ok=True)
with open(args.dst_save_pth, 'wb') as fp:
pickle.dump((img_paths, np.concatenate(all_features), np.concatenate(all_y)),
fp, pickle.HIGHEST_PROTOCOL)
print('Final batch:')
progress.display(len(test_loader))
if args.is_root:
print('Results:')
print(','.join([
f'acc1={acc1.avg}',
f'acc10={acc10.avg}',
f'acc20={acc20.avg}',
f'precision1={precision1.avg}',
f'precision5={precision5.avg}',
f'precision10={precision10.avg}',
f'precision15={precision15.avg}',
f'precision20={precision20.avg}'
]))
if __name__ == '__main__':
parser = get_argparse()
main(parser.parse_args())