-
Notifications
You must be signed in to change notification settings - Fork 2
/
dividemix_clothing1m_loader.py
158 lines (132 loc) · 7.7 KB
/
dividemix_clothing1m_loader.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
import os
import random
import numpy as np
import json
import torch
import torchvision.transforms as transforms
from torch.utils.data import Dataset, DataLoader
from PIL import Image
class clothing_dataset(Dataset):
def __init__(self, root, transform, mode, num_samples=0, pred=[], probability=[], paths=[], num_class=14, train_targets=None, use_noisy_val=False):
self.mode = mode
self.transform = transform
if not use_noisy_val: # benchmark setting
trainfile = os.path.join(root, "annotations/noisy_train.txt")
valfile = os.path.join(root, "annotations/clean_val.txt")
testfile = os.path.join(root, "annotations/clean_test.txt")
else: # using a nnoisy validation setting, saving clean labels for training
trainfile = os.path.join(root, "noisy_val_annotations/nv_noisy_train.txt")
valfile = os.path.join(root, "noisy_val_annotations/nv_noisy_val.txt")
testfile = os.path.join(root, "noisy_val_annotations/nv_clean_test.txt")
# if given training targets, e.g., distill targets from a teacher model
# Training impaths and targets. The map is necessary since only images paths are given for the inilization of 'labeled subset'
impaths, targets = [], []
with open(trainfile,'r') as f:
for line in f.readlines():
row = line.split(" ")
impaths.append(root + '/' + row[0])
targets.append(int(row[1]))
if train_targets is not None:
targets = train_targets
self.train_label_map = {impaths[i]:targets[i] for i in range(len(impaths))}
# Randomly sampling class-balanced samples from the whole the training set (at each epoch)
if mode == 'all':
indexs = np.arange(len(impaths))
random.shuffle(indexs)
class_num = torch.zeros(num_class) # number of samples in each class
self.impaths, self.targets = [], []
for i in indexs:
label = np.argmax(targets[i]) if isinstance(targets[i], (list, np.ndarray)) else targets[i]
if class_num[label]<(num_samples/14) and len(self.impaths)<num_samples:
self.impaths.append(impaths[i])
self.targets.append(targets[i])
class_num[label]+=1
# trusted labeled subset
elif mode == 'labeled':
impaths = paths
pred_idx = pred.nonzero()[0]
self.impaths = [impaths[i] for i in pred_idx]
self.probability = [probability[i] for i in pred_idx]
self.targets = [self.train_label_map[img_path] for img_path in self.impaths]
print('{} data has a size of {}'.format(self.mode,len(self.impaths)))
# unlabeled subset
elif mode == 'unlabeled':
impaths = paths
pred_idx = (1-pred).nonzero()[0]
self.impaths = [impaths[i] for i in pred_idx]
self.probability = [probability[i] for i in pred_idx]
print('{} data has a size of {}'.format(self.mode,len(self.impaths)))
# val/test
elif mode in ['val', 'test']:
self.impaths, self.targets = [], []
flist = valfile if mode=='val' else testfile
with open(flist, 'r') as f:
for line in f.readlines():
row = line.split(" ")
self.impaths.append(root + '/' + row[0])
self.targets.append(int(row[1]))
def __getitem__(self, index):
if self.mode=='all':
img_path = self.impaths[index]
img = Image.open(img_path).convert('RGB')
img = self.transform(img)
target = self.targets[index]
return img, target, img_path
elif self.mode=='labeled':
img = Image.open(self.impaths[index]).convert('RGB')
img1 = self.transform(img)
img2 = self.transform(img)
target = self.targets[index]
prob = self.probability[index]
return img1, img2, target, prob
elif self.mode=='unlabeled':
img = Image.open(self.impaths[index]).convert('RGB')
img1 = self.transform(img)
img2 = self.transform(img)
return img1, img2
elif self.mode in ['val', 'test']:
img = Image.open(self.impaths[index]).convert('RGB')
img = self.transform(img)
target = self.targets[index]
return img, target
def __len__(self):
return len(self.impaths)
class clothing_dataloader():
def __init__(self, root, batch_size, num_batches, num_workers, use_noisy_val=False):
self.batch_size = batch_size
self.num_workers = num_workers
self.num_batches = num_batches
self.root = root
self.use_noisy_val = use_noisy_val
self.transform_train = transforms.Compose([
transforms.Resize(256),
transforms.RandomCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.6959, 0.6537, 0.6371),(0.3113, 0.3192, 0.3214)),
])
self.transform_test = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize((0.6959, 0.6537, 0.6371),(0.3113, 0.3192, 0.3214)),
])
def run(self,mode,pred=[],prob=[],paths=[], train_targets=None):
if mode=='warmup':
warmup_dataset = clothing_dataset(self.root,transform=self.transform_train, mode='all',num_samples=self.num_batches*self.batch_size*2,train_targets=train_targets, use_noisy_val=self.use_noisy_val)
warmup_loader = DataLoader(dataset=warmup_dataset, batch_size=self.batch_size*2, shuffle=True, num_workers=self.num_workers)
return warmup_loader
elif mode=='train':
labeled_dataset = clothing_dataset(self.root,transform=self.transform_train, mode='labeled',pred=pred, probability=prob,paths=paths,train_targets=train_targets, use_noisy_val=self.use_noisy_val)
labeled_loader = DataLoader(dataset=labeled_dataset, batch_size=self.batch_size, shuffle=True, num_workers=self.num_workers)
unlabeled_dataset = clothing_dataset(self.root,transform=self.transform_train, mode='unlabeled',pred=pred, probability=prob,paths=paths, use_noisy_val=self.use_noisy_val)
unlabeled_loader = DataLoader(dataset=unlabeled_dataset, batch_size=self.batch_size, shuffle=True, num_workers=self.num_workers)
return labeled_loader,unlabeled_loader
elif mode=='eval_train':
eval_dataset = clothing_dataset(self.root,transform=self.transform_test, mode='all', num_samples=self.num_batches*self.batch_size,train_targets=train_targets, use_noisy_val=self.use_noisy_val)
eval_loader = DataLoader(dataset=eval_dataset, batch_size=self.batch_size, shuffle=False, num_workers=self.num_workers)
return eval_loader
elif mode in ['val', 'test']:
dataset = clothing_dataset(self.root,transform=self.transform_test, mode=mode, use_noisy_val=self.use_noisy_val)
loader = DataLoader(dataset=dataset, batch_size=128, shuffle=False, num_workers=self.num_workers)
return loader