-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathcv_utils.py
357 lines (297 loc) · 14.9 KB
/
cv_utils.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
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
#
# Copyright (c) 2023 salesforce.com, inc.
# All rights reserved.
# SPDX-License-Identifier: Apache-2.0
# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/Apache-2.0
#
"""
Utilities for computer vision experiments.
"""
import hashlib
import logging
import os
import tarfile
from datasets import load_dataset
import numpy as np
import requests
import torch
import torch.distributed as dist
import torch.nn as nn
from torch.utils.data import Dataset
import torchvision
from torchvision.datasets import ImageFolder
import torchvision.transforms as transforms
import tqdm
logger = logging.getLogger(__name__)
def create_model(dataset, model_name="resnet18", device=torch.device("cpu"), **kwargs):
"""
Returns a model with a representation pre-trained on ImageNet, but an untrained final linear classification layer.
"""
model_name = model_name.lower()
if model_name == "resnet18":
model = torchvision.models.resnet18(weights="DEFAULT", **kwargs)
if not isinstance(dataset, ImageNet):
model.fc = nn.Linear(512, dataset.n_class)
elif model_name == "resnet50":
model = torchvision.models.resnet50(weights="DEFAULT", **kwargs)
if not isinstance(dataset, ImageNet):
model.fc = nn.Linear(2048, dataset.n_class)
elif model_name == "densenet121":
model = torchvision.models.densenet121(weights="DEFAULT", **kwargs)
if not isinstance(dataset, ImageNet):
model.classifier = nn.Linear(1024, dataset.n_class)
elif model_name == "inception_v3":
model = torchvision.models.inception_v3(weights="DEFAULT", transform_input=False, **kwargs)
if not isinstance(dataset, ImageNet):
model.fc = nn.Linear(2048, dataset.n_class)
elif model_name == "wide_resnet50":
model = torchvision.models.wide_resnet50_2(weights="DEFAULT", **kwargs)
if not isinstance(dataset, ImageNet):
model.fc = nn.Linear(2048, dataset.n_class)
else:
raise NotImplementedError(f"Model {model_name} is not a supported pre-trained ImageNet model.")
return model.to(device=device)
def data_loader(dataset, batch_size=256, epoch=0, pin_memory=True):
shuffle = dataset.split == "train"
if dist.is_available() and dist.is_initialized():
sampler = torch.utils.data.DistributedSampler(dataset, shuffle=shuffle)
sampler.set_epoch(epoch)
elif shuffle:
sampler = torch.utils.data.RandomSampler(dataset)
else:
sampler = torch.utils.data.SequentialSampler(dataset)
return torch.utils.data.DataLoader(dataset, batch_size=batch_size, sampler=sampler, pin_memory=pin_memory)
class DatasetMixIn(Dataset):
def __init__(self, split):
assert split in ["train", "valid", "test"]
self.split = split
if split == "train":
resize = [transforms.Resize(256), transforms.RandomResizedCrop(224)]
else:
resize = [transforms.Resize(256), transforms.CenterCrop(224)]
self.transform = transforms.Compose([*resize, transforms.ToTensor(), self.norm()])
self.instance_key = "image"
self.label_key = "label"
def __len__(self):
test_name = "test" if "test" in self.data else "valid"
if self.split == "train":
return len(self.train_idx)
if self.split == "valid":
return len(self.valid_idx)
return len(self.data[test_name])
def __getitem__(self, i):
test_name = "test" if "test" in self.data else "valid"
if self.split == "train":
i = self.train_idx[i]
elif self.split == "valid":
i = self.valid_idx[i]
# Get the data point
instance = self.data[test_name if self.split == "test" else "train"][i]
return self.transform(instance[self.instance_key].convert("RGB")), instance[self.label_key]
class ImageNet:
def __init__(self, split, rootdir="/export/share/datasets/vision/imagenet"):
if split == "train":
self.data = ImageFolder(os.path.join(rootdir, "train"))
else:
self.data = ImageFolder(os.path.join(rootdir, "val"))
resize = [transforms.Resize(256), transforms.CenterCrop(224)]
self.split = split
self.transform = transforms.Compose([*resize, transforms.ToTensor(), TinyImageNet.norm()])
def __len__(self):
return len(self.data) if self.split == "train" else len(self.data) // 2
def __getitem__(self, i):
img, label = self.data[i if self.split == "train" else 2 * i + (self.split == "test")]
return self.transform(img), label
@property
def n_class(self):
return 1000
class TinyImageNet(DatasetMixIn):
def __init__(self, split):
super().__init__(split)
self.data = load_dataset("Maysee/tiny-imagenet")
self.valid_idx = list(range(0, 100000, 10))
self.train_idx = [i for i in range(100000) if i not in self.valid_idx]
@classmethod
def norm(cls):
return transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
@property
def n_class(self):
return 200
class CIFAR10(DatasetMixIn):
def __init__(self, split):
super().__init__(split)
self.data = load_dataset("cifar10")
self.train_idx = range(45000)
self.valid_idx = range(45000, 50000)
self.instance_key = "img"
@classmethod
def norm(cls):
return transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
@property
def n_class(self):
return 10
class CIFAR100(DatasetMixIn):
def __init__(self, split):
super().__init__(split)
self.data = load_dataset("cifar100")
self.train_idx = range(45000)
self.valid_idx = range(45000, 50000)
self.instance_key = "img"
self.label_key = "fine_label"
@classmethod
def norm(cls):
return transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
@property
def n_class(self):
return 100
class Downloader(Dataset):
def __len__(self):
return len(self.data)
def __getitem__(self, i):
image, label = self.data[i]
if not isinstance(self.data, DatasetMixIn):
image = self.transform(image)
return image, label
@staticmethod
def extract_tar(tar_path, extract_dir=None, success_file=None):
if extract_dir is None:
extract_dir = os.path.join(os.path.dirname(tar_path), os.path.basename(tar_path)[: -len(".tar")])
success_file = os.path.join(extract_dir, "_SUCCESS") if success_file is None else success_file
if not os.path.isfile(success_file):
logger.info(f"Extracting tarfile {os.path.basename(tar_path)}...")
tarfile.open(tar_path).extractall(path=os.path.dirname(tar_path))
with open(success_file, "w"):
pass
@staticmethod
def download(url, file_name, expected_md5):
if os.path.exists(file_name):
logger.info("Checking MD5 checksum...")
with open(file_name, "rb") as file_to_check:
data = file_to_check.read()
md5_returned = hashlib.md5(data).hexdigest()
if md5_returned != expected_md5:
logger.info("Invalid MD5 checksum. Restarting download.")
os.remove(file_name)
if os.path.exists(file_name):
return
logger.info(f"Downloading file {os.path.basename(file_name)}...")
os.makedirs(os.path.dirname(file_name), exist_ok=True)
with open(file_name, "wb") as f:
r = requests.get(url, stream=True)
bar_format = "{l_bar}{bar}| {n:.1f}/{total:.1f}MB [{elapsed}<{remaining}, {rate_fmt}]"
total_mb = int(r.headers["Content-Length"]) / (1024**2)
with tqdm.tqdm(unit="MB", total=total_mb, bar_format=bar_format) as pbar:
for chunk in r.iter_content(chunk_size=1 * (1024**2)):
if chunk: # filter out keep-alive new chunks
f.write(chunk)
f.flush()
pbar.update(len(chunk) / (1024**2))
class ImageNetC(Downloader):
def __init__(self, corruption=None, severity=0):
# Download dataset
base_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "data", "ImageNetC")
if not os.path.exists(os.path.join(base_dir, "_SUCCESS")):
base = "https://zenodo.org/record/2235448/files"
suf = "?download=1"
self.download(f"{base}/blur.tar{suf}", f"{base_dir}/blur.tar", "2d8e81fdd8e07fef67b9334fa635e45c")
self.download(f"{base}/digital.tar{suf}", f"{base_dir}/digital.tar", "89157860d7b10d5797849337ca2e5c03")
self.download(f"{base}/noise.tar{suf}", f"{base_dir}/noise.tar", "e80562d7f6c3f8834afb1ecf27252745")
self.download(f"{base}/weather.tar{suf}", f"{base_dir}/weather.tar", "33ffea4db4d93fe4a428c40a6ce0c25d")
with open(os.path.join(base_dir, "_SUCCESS"), "w"):
pass
# Extract tar files
self.extract_tar(f"{base_dir}/blur.tar", success_file=f"{base_dir}/_SUCCESS_blur")
self.extract_tar(f"{base_dir}/digital.tar", success_file=f"{base_dir}/_SUCCESS_digital")
self.extract_tar(f"{base_dir}/noise.tar", success_file=f"{base_dir}/_SUCCESS_noise")
self.extract_tar(f"{base_dir}/weather.tar", success_file=f"{base_dir}/_SUCCESS_weather")
# Get the actual dataset
if severity == 0 or corruption is None:
self.data = ImageNet(split="test")
else:
valid_corruptions = [d for d in os.listdir(base_dir) if os.path.isdir(os.path.join(base_dir, d))]
assert severity in range(1, 6), f"Got severity={severity}. Expected an int between 1 and 5."
assert corruption in valid_corruptions, f"Got corruption={corruption}. Expected one of {valid_corruptions}"
self.data = ImageFolder(os.path.join(base_dir, corruption, str(severity)))
resize = [transforms.Resize(256), transforms.CenterCrop(224)]
self.transform = transforms.Compose([*resize, transforms.ToTensor(), TinyImageNet.norm()])
self.split = "test"
def __getitem__(self, i):
return self.data[i] if isinstance(self.data, ImageNet) else super().__getitem__(i)
@property
def n_class(self):
return 1000
class TinyImageNetC(Downloader):
def __init__(self, corruption=None, severity=0):
# Download data & extract tar if needed
url = "https://zenodo.org/record/2469796/files/TinyImageNet-C.tar?download=1"
base_dir = os.path.dirname(os.path.abspath(__file__))
file_name = os.path.join(base_dir, "data", "TinyImageNet-C.tar")
data_dir = os.path.join(os.path.dirname(file_name), "TinyImageNet-C", "Tiny-ImageNet-C")
if not os.path.exists(os.path.join(data_dir, "_SUCCESS")):
self.download(url=url, file_name=file_name, expected_md5="3d9c6e89c2609aeb4198f84c8edd1ff0")
self.extract_tar(file_name)
self.extract_tar(os.path.join(os.path.dirname(file_name), "TinyImageNet-C", "Tiny-ImageNet-C.tar"))
# Get the actual dataset
if severity == 0 or corruption is None:
self.data = TinyImageNet(split="test")
else:
valid_corruptions = [d for d in os.listdir(data_dir) if os.path.isdir(os.path.join(data_dir, d))]
assert severity in range(1, 6), f"Got severity={severity}. Expected an int between 1 and 5."
assert corruption in valid_corruptions, f"Got corruption={corruption}. Expected one of {valid_corruptions}"
self.data = ImageFolder(os.path.join(data_dir, corruption, str(severity)))
resize = [transforms.Resize(256), transforms.CenterCrop(224)]
self.transform = transforms.Compose([*resize, transforms.ToTensor(), TinyImageNet.norm()])
self.split = "test"
@property
def n_class(self):
return 200
class CIFAR10C(Downloader):
def __init__(self, corruption=None, severity=0):
# Download data & extract tar if needed
url = "https://zenodo.org/record/2535967/files/CIFAR-10-C.tar?download=1"
base_dir = os.path.dirname(os.path.abspath(__file__))
file_name = os.path.join(base_dir, "data", "CIFAR10-C.tar")
data_dir = os.path.join(base_dir, "data", "CIFAR-10-C")
if not os.path.exists(os.path.join(data_dir, "_SUCCESS")):
self.download(url=url, file_name=file_name, expected_md5="56bf5dcef84df0e2308c6dcbcbbd8499")
self.extract_tar(file_name, extract_dir=data_dir)
# Get the actual dataset
if severity == 0 or corruption is None:
self.data = CIFAR10(split="test")
else:
valid_corruptions = [d[:-4] for d in os.listdir(data_dir) if d != "labels.npy" and d.endswith(".npy")]
assert severity in range(1, 6), f"Got severity={severity}. Expected an int between 1 and 5."
assert corruption in valid_corruptions, f"Got corruption={corruption}. Expected one of {valid_corruptions}"
data = np.load(os.path.join(data_dir, corruption + ".npy"))
labels = np.load(os.path.join(data_dir, "labels.npy"))
self.data = [(data[i], labels[i]) for i in range((severity - 1) * 10000, severity * 10000)]
self.transform = transforms.Compose([transforms.ToTensor(), transforms.Resize(224), CIFAR10.norm()])
self.split = "test"
@property
def n_class(self):
return 10
class CIFAR100C(Downloader):
def __init__(self, corruption=None, severity=0):
# Download data & extract tar if needed
url = "https://zenodo.org/record/3555552/files/CIFAR-100-C.tar?download=1"
base_dir = os.path.dirname(os.path.abspath(__file__))
file_name = os.path.join(base_dir, "data", "CIFAR100-C.tar")
data_dir = os.path.join(base_dir, "data", "CIFAR-100-C")
if not os.path.exists(os.path.join(data_dir, "_SUCCESS")):
self.download(url=url, file_name=file_name, expected_md5="11f0ed0f1191edbf9fa23466ae6021d3")
self.extract_tar(file_name, extract_dir=data_dir)
# Get the actual dataset
if severity == 0 or corruption is None:
self.data = CIFAR100(split="test")
else:
valid_corruptions = [d[:-4] for d in os.listdir(data_dir) if d != "labels.npy" and d.endswith(".npy")]
assert severity in range(1, 6), f"Got severity={severity}. Expected an int between 1 and 5."
assert corruption in valid_corruptions, f"Got corruption={corruption}. Expected one of {valid_corruptions}"
data = np.load(os.path.join(data_dir, corruption + ".npy"))
labels = np.load(os.path.join(data_dir, "labels.npy"))
self.data = [(data[i], labels[i]) for i in range((severity - 1) * 10000, severity * 10000)]
self.transform = transforms.Compose([transforms.ToTensor(), transforms.Resize(224), CIFAR10.norm()])
self.split = "test"
@property
def n_class(self):
return 100