Spaces:
Configuration error
Configuration error
File size: 11,098 Bytes
72fc481 |
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 sys
import numpy as np
from PIL import Image
import torchvision
from torch.utils.data.dataset import Subset
from sklearn.metrics.pairwise import cosine_similarity, euclidean_distances
import torch
import torch.nn.functional as F
import random
import os
import json
from numpy.testing import assert_array_almost_equal
def get_cifar100(root, cfg_trainer, train=True,
transform_train=None, transform_val=None,
download=False, noise_file = ''):
base_dataset = torchvision.datasets.CIFAR100(root, train=train, download=download)
if train:
train_idxs, val_idxs = train_val_split(base_dataset.targets)
train_dataset = CIFAR100_train(root, cfg_trainer, train_idxs, train=True, transform=transform_train)
val_dataset = CIFAR100_val(root, cfg_trainer, val_idxs, train=train, transform=transform_val)
if cfg_trainer['asym']:
train_dataset.asymmetric_noise()
val_dataset.asymmetric_noise()
else:
train_dataset.symmetric_noise()
val_dataset.symmetric_noise()
print(f"Train: {len(train_dataset)} Val: {len(val_dataset)}") # Train: 45000 Val: 5000
else:
train_dataset = []
val_dataset = CIFAR100_val(root, cfg_trainer, None, train=train, transform=transform_val)
print(f"Test: {len(val_dataset)}")
return train_dataset, val_dataset
def train_val_split(base_dataset: torchvision.datasets.CIFAR100):
num_classes = 100
base_dataset = np.array(base_dataset)
train_n = int(len(base_dataset) * 0.9 / num_classes)
train_idxs = []
val_idxs = []
for i in range(num_classes):
idxs = np.where(base_dataset == i)[0]
np.random.shuffle(idxs)
train_idxs.extend(idxs[:train_n])
val_idxs.extend(idxs[train_n:])
np.random.shuffle(train_idxs)
np.random.shuffle(val_idxs)
return train_idxs, val_idxs
class CIFAR100_train(torchvision.datasets.CIFAR100):
def __init__(self, root, cfg_trainer, indexs, train=True,
transform=None, target_transform=None,
download=False):
super(CIFAR100_train, self).__init__(root, train=train,
transform=transform, target_transform=target_transform,
download=download)
self.num_classes = 100
self.cfg_trainer = cfg_trainer
self.train_data = self.data[indexs]
self.train_labels = np.array(self.targets)[indexs]
self.indexs = indexs
self.prediction = np.zeros((len(self.train_data), self.num_classes, self.num_classes), dtype=np.float32)
self.noise_indx = []
#self.all_refs_encoded = torch.zeros(self.num_classes,self.num_ref,1024, dtype=np.float32)
self.count = 0
def symmetric_noise(self):
self.train_labels_gt = self.train_labels.copy()
indices = np.random.permutation(len(self.train_data))
for i, idx in enumerate(indices):
if i < self.cfg_trainer['percent'] * len(self.train_data):
self.noise_indx.append(idx)
self.train_labels[idx] = np.random.randint(self.num_classes, dtype=np.int32)
def multiclass_noisify(self, y, P, random_state=0):
""" Flip classes according to transition probability matrix T.
It expects a number between 0 and the number of classes - 1.
"""
assert P.shape[0] == P.shape[1]
assert np.max(y) < P.shape[0]
# row stochastic matrix
assert_array_almost_equal(P.sum(axis=1), np.ones(P.shape[1]))
assert (P >= 0.0).all()
m = y.shape[0]
new_y = y.copy()
flipper = np.random.RandomState(random_state)
for idx in np.arange(m):
i = y[idx]
# draw a vector with only an 1
flipped = flipper.multinomial(1, P[i, :], 1)[0]
new_y[idx] = np.where(flipped == 1)[0]
return new_y
# def build_for_cifar100(self, size, noise):
# """ random flip between two random classes.
# """
# assert(noise >= 0.) and (noise <= 1.)
# P = np.eye(size)
# cls1, cls2 = np.random.choice(range(size), size=2, replace=False)
# P[cls1, cls2] = noise
# P[cls2, cls1] = noise
# P[cls1, cls1] = 1.0 - noise
# P[cls2, cls2] = 1.0 - noise
# assert_array_almost_equal(P.sum(axis=1), 1, 1)
# return P
def build_for_cifar100(self, size, noise):
""" The noise matrix flips to the "next" class with probability 'noise'.
"""
assert(noise >= 0.) and (noise <= 1.)
P = (1. - noise) * np.eye(size)
for i in np.arange(size - 1):
P[i, i + 1] = noise
# adjust last row
P[size - 1, 0] = noise
assert_array_almost_equal(P.sum(axis=1), 1, 1)
return P
def asymmetric_noise(self, asym=False, random_shuffle=False):
self.train_labels_gt = self.train_labels.copy()
P = np.eye(self.num_classes)
n = self.cfg_trainer['percent']
nb_superclasses = 20
nb_subclasses = 5
if n > 0.0:
for i in np.arange(nb_superclasses):
init, end = i * nb_subclasses, (i+1) * nb_subclasses
P[init:end, init:end] = self.build_for_cifar100(nb_subclasses, n)
y_train_noisy = self.multiclass_noisify(self.train_labels, P=P,
random_state=0)
actual_noise = (y_train_noisy != self.train_labels).mean()
assert actual_noise > 0.0
self.train_labels = y_train_noisy
def __getitem__(self, index):
"""
Args:
index (int): Index
Returns:
tuple: (image, target) where target is index of the target class.
"""
img, target, target_gt = self.train_data[index], self.train_labels[index], self.train_labels_gt[index]
# doing this so that it is consistent with all other datasets
# to return a PIL Image
img = Image.fromarray(img)
if self.transform is not None:
img = self.transform(img)
if self.target_transform is not None:
target = self.target_transform(target)
return img, target, index, target_gt
def __len__(self):
return len(self.train_data)
class CIFAR100_val(torchvision.datasets.CIFAR100):
def __init__(self, root, cfg_trainer, indexs, train=True,
transform=None, target_transform=None,
download=False):
super(CIFAR100_val, self).__init__(root, train=train,
transform=transform, target_transform=target_transform,
download=download)
# self.train_data = self.data[indexs]
# self.train_labels = np.array(self.targets)[indexs]
self.num_classes = 100
self.cfg_trainer = cfg_trainer
if train:
self.train_data = self.data[indexs]
self.train_labels = np.array(self.targets)[indexs]
else:
self.train_data = self.data
self.train_labels = np.array(self.targets)
self.train_labels_gt = self.train_labels.copy()
def symmetric_noise(self):
indices = np.random.permutation(len(self.train_data))
for i, idx in enumerate(indices):
if i < self.cfg_trainer['percent'] * len(self.train_data):
self.train_labels[idx] = np.random.randint(self.num_classes, dtype=np.int32)
def multiclass_noisify(self, y, P, random_state=0):
""" Flip classes according to transition probability matrix T.
It expects a number between 0 and the number of classes - 1.
"""
assert P.shape[0] == P.shape[1]
assert np.max(y) < P.shape[0]
# row stochastic matrix
assert_array_almost_equal(P.sum(axis=1), np.ones(P.shape[1]))
assert (P >= 0.0).all()
m = y.shape[0]
new_y = y.copy()
flipper = np.random.RandomState(random_state)
for idx in np.arange(m):
i = y[idx]
# draw a vector with only an 1
flipped = flipper.multinomial(1, P[i, :], 1)[0]
new_y[idx] = np.where(flipped == 1)[0]
return new_y
# def build_for_cifar100(self, size, noise):
# """ random flip between two random classes.
# """
# assert(noise >= 0.) and (noise <= 1.)
# P = np.eye(size)
# cls1, cls2 = np.random.choice(range(size), size=2, replace=False)
# P[cls1, cls2] = noise
# P[cls2, cls1] = noise
# P[cls1, cls1] = 1.0 - noise
# P[cls2, cls2] = 1.0 - noise
# assert_array_almost_equal(P.sum(axis=1), 1, 1)
# return P
def build_for_cifar100(self, size, noise):
""" The noise matrix flips to the "next" class with probability 'noise'.
"""
assert(noise >= 0.) and (noise <= 1.)
P = (1. - noise) * np.eye(size)
for i in np.arange(size - 1):
P[i, i + 1] = noise
# adjust last row
P[size - 1, 0] = noise
assert_array_almost_equal(P.sum(axis=1), 1, 1)
return P
def asymmetric_noise(self, asym=False, random_shuffle=False):
P = np.eye(self.num_classes)
n = self.cfg_trainer['percent']
nb_superclasses = 20
nb_subclasses = 5
if n > 0.0:
for i in np.arange(nb_superclasses):
init, end = i * nb_subclasses, (i+1) * nb_subclasses
P[init:end, init:end] = self.build_for_cifar100(nb_subclasses, n)
y_train_noisy = self.multiclass_noisify(self.train_labels, P=P,
random_state=0)
actual_noise = (y_train_noisy != self.train_labels).mean()
assert actual_noise > 0.0
self.train_labels = y_train_noisy
def __len__(self):
return len(self.train_data)
def __getitem__(self, index):
"""
Args:
index (int): Index
Returns:
tuple: (image, target) where target is index of the target class.
"""
img, target, target_gt = self.train_data[index], self.train_labels[index], self.train_labels_gt[index]
# doing this so that it is consistent with all other datasets
# to return a PIL Image
img = Image.fromarray(img)
if self.transform is not None:
img = self.transform(img)
if self.target_transform is not None:
target = self.target_transform(target)
return img, target, index, target_gt
|