-
Notifications
You must be signed in to change notification settings - Fork 2.7k
Expand file tree
/
Copy pathdatasets.py
More file actions
169 lines (146 loc) · 6.35 KB
/
datasets.py
File metadata and controls
169 lines (146 loc) · 6.35 KB
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
# --------------------------------------------------------
# BEIT: BERT Pre-Training of Image Transformers (https://arxiv.org/abs/2106.08254)
# Github source: https://github.com/microsoft/unilm/tree/master/beit
# Copyright (c) 2021 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# By Hangbo Bao
# Based on timm, DINO and DeiT code bases
# https://github.com/rwightman/pytorch-image-models/tree/master/timm
# https://github.com/facebookresearch/deit/
# https://github.com/facebookresearch/dino
# --------------------------------------------------------'
import os
import torch
from torchvision import datasets, transforms
from timm.data.constants import \
IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD, IMAGENET_INCEPTION_MEAN, IMAGENET_INCEPTION_STD
from transforms import RandomResizedCropAndInterpolationWithTwoPic
from timm.data import create_transform
from dall_e.utils import map_pixels
from masking_generator import MaskingGenerator
from dataset_folder import ImageFolder
class DataAugmentationForBEiT(object):
def __init__(self, args):
imagenet_default_mean_and_std = args.imagenet_default_mean_and_std
mean = IMAGENET_INCEPTION_MEAN if not imagenet_default_mean_and_std else IMAGENET_DEFAULT_MEAN
std = IMAGENET_INCEPTION_STD if not imagenet_default_mean_and_std else IMAGENET_DEFAULT_STD
self.common_transform = transforms.Compose([
transforms.ColorJitter(0.4, 0.4, 0.4),
transforms.RandomHorizontalFlip(p=0.5),
RandomResizedCropAndInterpolationWithTwoPic(
size=args.input_size, second_size=args.second_input_size,
interpolation=args.train_interpolation, second_interpolation=args.second_interpolation,
),
])
self.patch_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(
mean=torch.tensor(mean),
std=torch.tensor(std))
])
if args.discrete_vae_type == "dall-e":
self.visual_token_transform = transforms.Compose([
transforms.ToTensor(),
map_pixels,
])
elif args.discrete_vae_type == "customized":
self.visual_token_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(
mean=IMAGENET_INCEPTION_MEAN,
std=IMAGENET_INCEPTION_STD,
),
])
else:
raise NotImplementedError()
self.masked_position_generator = MaskingGenerator(
args.window_size, num_masking_patches=args.num_mask_patches,
max_num_patches=args.max_mask_patches_per_block,
min_num_patches=args.min_mask_patches_per_block,
)
def __call__(self, image):
for_patches, for_visual_tokens = self.common_transform(image)
return \
self.patch_transform(for_patches), self.visual_token_transform(for_visual_tokens), \
self.masked_position_generator()
def __repr__(self):
repr = "(DataAugmentationForBEiT,\n"
repr += " common_transform = %s,\n" % str(self.common_transform)
repr += " patch_transform = %s,\n" % str(self.patch_transform)
repr += " visual_tokens_transform = %s,\n" % str(self.visual_token_transform)
repr += " Masked position generator = %s,\n" % str(self.masked_position_generator)
repr += ")"
return repr
def build_beit_pretraining_dataset(args):
transform = DataAugmentationForBEiT(args)
print("Data Aug = %s" % str(transform))
return ImageFolder(args.data_path, transform=transform)
def build_dataset(is_train, args):
transform = build_transform(is_train, args)
print("Transform = ")
if isinstance(transform, tuple):
for trans in transform:
print(" - - - - - - - - - - ")
for t in trans.transforms:
print(t)
else:
for t in transform.transforms:
print(t)
print("---------------------------")
if args.data_set == 'CIFAR':
dataset = datasets.CIFAR100(args.data_path, train=is_train, transform=transform)
nb_classes = 100
elif args.data_set == 'IMNET':
root = os.path.join(args.data_path, 'train' if is_train else 'val')
dataset = datasets.ImageFolder(root, transform=transform)
nb_classes = 1000
elif args.data_set == "image_folder":
root = args.data_path if is_train else args.eval_data_path
dataset = ImageFolder(root, transform=transform)
nb_classes = args.nb_classes
assert len(dataset.class_to_idx) == nb_classes
else:
raise NotImplementedError()
assert nb_classes == args.nb_classes
print("Number of the class = %d" % args.nb_classes)
return dataset, nb_classes
def build_transform(is_train, args):
resize_im = args.input_size > 32
imagenet_default_mean_and_std = args.imagenet_default_mean_and_std
mean = IMAGENET_INCEPTION_MEAN if not imagenet_default_mean_and_std else IMAGENET_DEFAULT_MEAN
std = IMAGENET_INCEPTION_STD if not imagenet_default_mean_and_std else IMAGENET_DEFAULT_STD
if is_train:
# this should always dispatch to transforms_imagenet_train
transform = create_transform(
input_size=args.input_size,
is_training=True,
color_jitter=args.color_jitter,
auto_augment=args.aa,
interpolation=args.train_interpolation,
re_prob=args.reprob,
re_mode=args.remode,
re_count=args.recount,
mean=mean,
std=std,
)
if not resize_im:
# replace RandomResizedCropAndInterpolation with
# RandomCrop
transform.transforms[0] = transforms.RandomCrop(
args.input_size, padding=4)
return transform
t = []
if resize_im:
if args.crop_pct is None:
if args.input_size < 384:
args.crop_pct = 224 / 256
else:
args.crop_pct = 1.0
size = int(args.input_size / args.crop_pct)
t.append(
transforms.Resize(size, interpolation=3), # to maintain same ratio w.r.t. 224 images
)
t.append(transforms.CenterCrop(args.input_size))
t.append(transforms.ToTensor())
t.append(transforms.Normalize(mean, std))
return transforms.Compose(t)