Spaces:
Sleeping
Sleeping
File size: 9,065 Bytes
bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea 44066b7 bd421ea |
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 os
import torch
import torch.nn
import numpy as np
from PIL import Image
from skimage.io import imread
import torchvision.transforms as transforms
from torch.utils.data import Dataset, DataLoader
from sklearn.model_selection import train_test_split
def read_IAM_label_txt_file(file_txt_labels):
"""
---------
Arguments
---------
file_txt_labels : str
full path to the text file containing labels
-------
Returns
-------
a tuple of
all_image_files : list
a list of all image file names
all_labels : list
a list of all labels
"""
label_file_handler = open(file_txt_labels, mode="r")
all_lines = label_file_handler.readlines()
num_lines = len(all_lines)
all_image_files = []
all_labels = []
for cur_line_num in range(num_lines):
if cur_line_num % 3 == 0:
all_image_files.append(all_lines[cur_line_num].strip())
elif cur_line_num % 3 == 1:
all_labels.append(all_lines[cur_line_num].strip())
else:
continue
return all_image_files, all_labels
class HWRecogIAMDataset(Dataset):
"""
Main dataset class to be used only for training, validation and internal testing
"""
CHAR_SET = " !\"#&'()*+,-./0123456789:;?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
CHAR_2_LABEL = {char: i + 1 for i, char in enumerate(CHAR_SET)}
LABEL_2_CHAR = {label: char for char, label in CHAR_2_LABEL.items()}
def __init__(
self,
list_image_files,
list_labels,
dir_images,
image_height=32,
image_width=768,
which_set="train",
):
"""
---------
Arguments
---------
list_image_files : list
list of image files
list_labels : list
list of labels
dir_images : str
full path to directory containing images
image_height : int
image height (default: 32)
image_width : int
image width (default: 768)
which_set : str
a string indicating which set is being used (default: train)
"""
self.list_labels = list_labels
self.dir_images = dir_images
self.list_image_files = list_image_files
self.image_width = image_width
self.image_height = image_height
self.which_set = which_set
if self.which_set == "train":
# apply data augmentation only for train set
self.transform = transforms.Compose(
[
transforms.ToPILImage(),
transforms.Resize(
(self.image_height, self.image_width), Image.BILINEAR
),
transforms.RandomAffine(
degrees=[-0.75, 0.75],
translate=[0, 0.05],
scale=[0.75, 1],
shear=[-35, 35],
interpolation=transforms.InterpolationMode.BILINEAR,
fill=255,
),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225],
),
]
)
else:
self.transform = transforms.Compose(
[
transforms.ToPILImage(),
transforms.Resize(
(self.image_height, self.image_width), Image.BILINEAR
),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225],
),
]
)
def __len__(self):
return len(self.list_image_files)
def __getitem__(self, idx):
image_file_name = self.list_image_files[idx]
image_gray = imread(os.path.join(self.dir_images, image_file_name))
image_3_channel = np.repeat(np.expand_dims(image_gray, -1), 3, -1)
image_3_channel = self.transform(image_3_channel)
label_string = self.list_labels[idx]
label_encoded = [self.CHAR_2_LABEL[c] for c in label_string]
label_length = [len(label_encoded)]
label_encoded = torch.LongTensor(label_encoded)
label_length = torch.LongTensor(label_length)
return image_3_channel, label_encoded, label_length
def IAM_collate_fn(batch):
"""
collate function
---------
Arguments
---------
batch : tuple
a batch of input data as a tuple
-------
Returns
-------
a collated tuple of
images : tensor
tensor of batch images
labels : tensor
tensor of batch labels
label_lengths : tensor
tensor of batch label lengths
"""
images, labels, label_lengths = zip(*batch)
images = torch.stack(images, 0)
labels = torch.cat(labels, 0)
label_lengths = torch.cat(label_lengths, 0)
return images, labels, label_lengths
def split_dataset(file_txt_labels, for_train=True):
"""
---------
Arguments
---------
file_txt_labels : str
full path to the text file containing labels
for_train : bool
indicating whether split is for training or internal testing
-------
Returns
-------
a tuple of files depending for train or internal testing
"""
all_image_files, all_labels = read_IAM_label_txt_file(file_txt_labels)
train_image_files, test_image_files, train_labels, test_labels = train_test_split(
all_image_files, all_labels, test_size=0.1, random_state=4
)
train_image_files, valid_image_files, train_labels, valid_labels = train_test_split(
train_image_files, train_labels, test_size=0.1, random_state=4
)
if for_train:
return train_image_files, valid_image_files, train_labels, valid_labels
else:
return test_image_files, test_labels
def get_dataloaders_for_training(
train_x,
train_y,
valid_x,
valid_y,
dir_images,
image_height=32,
image_width=768,
batch_size=8,
):
"""
---------
Arguments
---------
train_x : list
list of train file names
train_y : list
list of train labels
valid_x : list
list of validation file names
valid_y : list
list of validation labels
dir_images : str
full directory path containing the images
image_height : int
image height (default: 32)
image_width : int
image width (default: 768)
batch_size : int
batch size (default: 8)
-------
Returns
-------
a tuple of dataloaders objects
train_loader : object
object of train set dataloader
valid_loader : object
object of validation set dataloader
"""
train_dataset = HWRecogIAMDataset(
train_x,
train_y,
dir_images,
image_height=image_height,
image_width=image_width,
which_set="train",
)
valid_dataset = HWRecogIAMDataset(
valid_x,
valid_y,
dir_images,
image_height=image_height,
image_width=image_width,
which_set="valid",
)
train_loader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=4,
collate_fn=IAM_collate_fn,
)
valid_loader = DataLoader(
valid_dataset,
batch_size=batch_size,
shuffle=False,
num_workers=4,
collate_fn=IAM_collate_fn,
)
return train_loader, valid_loader
def get_dataloader_for_testing(
test_x, test_y, dir_images, image_height=32, image_width=768, batch_size=1
):
"""
---------
Arguments
---------
test_x : list
list of test file names
test_y : list
list of test labels
dir_images : str
full directory path containing the images
image_height : int
image height (default: 32)
image_width : int
image width (default: 768)
batch_size : int
batch size (default: 1)
-------
Returns
-------
test_loader : object
object of test set dataloader
"""
test_dataset = HWRecogIAMDataset(
test_x,
test_y,
dir_images=dir_images,
image_height=image_height,
image_width=image_width,
which_set="test",
)
test_loader = DataLoader(
test_dataset,
batch_size=batch_size,
shuffle=False,
num_workers=4,
collate_fn=IAM_collate_fn,
)
return test_loader
|