Spaces:
Running on Zero
Running on Zero
File size: 13,557 Bytes
d500d65 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 | """Data loading and preprocessing utilities for classification and segmentation."""
import os
import random
from typing import List, Tuple, Optional
import numpy as np
import cv2
import yaml
import torch
from torch.utils.data import Dataset, DataLoader, Subset
from sklearn.model_selection import train_test_split
import albumentations as A
from albumentations.pytorch import ToTensorV2
def load_config(config_path: str = "config.yaml") -> dict:
"""Load configuration from a YAML file.
Args:
config_path: Path to the YAML configuration file.
Returns:
Dictionary containing configuration parameters.
"""
with open(config_path, "r", encoding="utf-8") as f:
return yaml.safe_load(f)
def set_seed(seed: int) -> None:
"""Set random seeds for reproducibility across libraries.
Args:
seed: Integer seed value.
"""
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
class ClassificationDataset(Dataset):
"""PyTorch Dataset for brain tumor classification.
Expects a directory structure where each class has its own subfolder
containing image files (JPG/PNG).
"""
def __init__(
self,
root_dir: str,
class_names: List[str],
img_size: int = 224,
transform: Optional[A.Compose] = None,
phase: str = "train",
):
"""Initialize the classification dataset.
Args:
root_dir: Root directory containing class subfolders.
class_names: Ordered list of class names.
img_size: Target image size (square).
transform: Albumentations composition to apply.
phase: Dataset phase identifier (train/val/test).
"""
self.root_dir = root_dir
self.class_names = class_names
self.img_size = img_size
self.transform = transform
self.phase = phase
self.samples: List[Tuple[str, int]] = []
self._build_samples()
def _build_samples(self) -> None:
"""Populate the samples list by scanning class directories."""
for idx, class_name in enumerate(self.class_names):
class_dir = os.path.join(self.root_dir, class_name)
if not os.path.isdir(class_dir):
continue
for fname in sorted(os.listdir(class_dir)):
if fname.lower().endswith((".png", ".jpg", ".jpeg")):
self.samples.append((os.path.join(class_dir, fname), idx))
def __len__(self) -> int:
"""Return the number of samples in the dataset."""
return len(self.samples)
def __getitem__(self, idx: int) -> Tuple[torch.Tensor, int]:
"""Retrieve a single image-label pair.
Args:
idx: Sample index.
Returns:
Tuple of transformed image tensor and integer label.
"""
img_path, label = self.samples[idx]
image = cv2.imread(img_path)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (self.img_size, self.img_size))
if self.transform is not None:
augmented = self.transform(image=image)
image = augmented["image"]
return image, label
class SegmentationDataset(Dataset):
"""PyTorch Dataset for brain tumor segmentation.
Recursively scans a directory for images and pairs them with masks
identified by a configurable suffix (e.g., image_mask.png).
"""
def __init__(
self,
image_dir: str,
mask_suffix: str = "_mask",
img_size: int = 128,
transform: Optional[A.Compose] = None,
):
"""Initialize the segmentation dataset.
Args:
image_dir: Root directory containing images and masks.
mask_suffix: Suffix identifying mask files.
img_size: Target image size (square).
transform: Albumentations composition to apply.
"""
self.image_dir = image_dir
self.mask_suffix = mask_suffix
self.img_size = img_size
self.transform = transform
self.pairs: List[Tuple[str, str]] = []
self._build_pairs()
def _build_pairs(self) -> None:
"""Populate image-mask pairs by scanning the directory tree."""
for root, _, files in os.walk(self.image_dir):
for fname in sorted(files):
if not fname.lower().endswith((".png", ".jpg", ".jpeg", ".tif", ".tiff")):
continue
if self.mask_suffix in fname:
continue
base, ext = os.path.splitext(fname)
mask_name = f"{base}{self.mask_suffix}{ext}"
mask_path = os.path.join(root, mask_name)
if os.path.exists(mask_path):
self.pairs.append((os.path.join(root, fname), mask_path))
def __len__(self) -> int:
"""Return the number of image-mask pairs."""
return len(self.pairs)
def __getitem__(self, idx: int) -> Tuple[torch.Tensor, torch.Tensor]:
"""Retrieve a single image-mask pair.
Args:
idx: Sample index.
Returns:
Tuple of transformed image tensor and binary mask tensor.
"""
img_path, mask_path = self.pairs[idx]
image = cv2.imread(img_path)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (self.img_size, self.img_size))
mask = cv2.imread(mask_path, cv2.IMREAD_GRAYSCALE)
mask = cv2.resize(mask, (self.img_size, self.img_size))
mask = (mask > 0).astype(np.float32)
if self.transform is not None:
augmented = self.transform(image=image, mask=mask)
image = augmented["image"]
mask = augmented["mask"].unsqueeze(0)
else:
image = torch.from_numpy(image.transpose(2, 0, 1)).float() / 255.0
mask = torch.from_numpy(mask).unsqueeze(0).float()
return image, mask
def get_classification_transforms(
img_size: int, augmentation: dict
) -> Tuple[A.Compose, A.Compose]:
"""Create Albumentations transforms for classification.
Args:
img_size: Target square image size.
augmentation: Augmentation parameters from config.
Returns:
Tuple of (train_transform, val_transform).
"""
extra = []
if augmentation.get("elastic_transform", 0) > 0:
extra.append(A.ElasticTransform(
alpha=1, sigma=50, p=0.5
))
if augmentation.get("grid_distortion", 0) > 0:
extra.append(A.GridDistortion(distort_limit=augmentation["grid_distortion"], p=0.5))
if augmentation.get("optical_distortion", 0) > 0:
extra.append(A.OpticalDistortion(
distort_limit=augmentation["optical_distortion"], p=0.5
))
if augmentation.get("gaussian_noise", 0) > 0:
extra.append(A.GaussNoise(std_range=(0.04, 0.20), p=0.5))
if augmentation.get("cutout", 0) > 0:
extra.append(A.CoarseDropout(
num_holes_range=(1, 8), hole_height_range=(0.0, 0.1), hole_width_range=(0.0, 0.1),
p=0.5
))
train_transform = A.Compose(
[
A.Resize(img_size, img_size),
A.HorizontalFlip(p=0.5)
if augmentation.get("random_flip") == "horizontal"
else A.NoOp(),
A.Rotate(
limit=int(augmentation["random_rotation"] * 180), p=0.5
),
A.RandomScale(
scale_limit=augmentation["random_zoom"], p=0.5
),
A.RandomBrightnessContrast(
brightness_limit=0, contrast_limit=augmentation["random_contrast"], p=0.5
),
]
+ extra
+ [
A.Resize(img_size, img_size),
A.Normalize(
mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)
),
ToTensorV2(),
]
)
val_transform = A.Compose(
[
A.Resize(img_size, img_size),
A.Normalize(
mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)
),
ToTensorV2(),
]
)
return train_transform, val_transform
def get_segmentation_transforms(img_size: int) -> Tuple[A.Compose, A.Compose]:
"""Create Albumentations transforms for segmentation.
Args:
img_size: Target square image size.
Returns:
Tuple of (train_transform, val_transform).
"""
train_transform = A.Compose(
[
A.Resize(img_size, img_size),
A.HorizontalFlip(p=0.5),
A.Rotate(limit=20, p=0.5),
A.RandomScale(scale_limit=0.1, p=0.5),
A.RandomBrightnessContrast(
brightness_limit=0, contrast_limit=0.1, p=0.5
),
A.Resize(img_size, img_size),
A.Normalize(
mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)
),
ToTensorV2(),
]
)
val_transform = A.Compose(
[
A.Resize(img_size, img_size),
A.Normalize(
mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)
),
ToTensorV2(),
]
)
return train_transform, val_transform
def get_classification_loaders(
config: dict, num_workers: int = 4
) -> Tuple[DataLoader, DataLoader, DataLoader, List[str]]:
"""Create train/validation/test data loaders for classification.
Args:
config: Loaded configuration dictionary.
num_workers: Number of workers for data loading.
Returns:
Tuple of (train_loader, val_loader, test_loader, class_names).
"""
set_seed(config["seed"])
data_dir = config["paths"]["data_classification"]
class_names = config["classification"]["class_names"]
img_size = config["classification"]["img_size"]
batch_size = config["classification"]["batch_size"]
aug = config["classification"]["augmentation"]
val_split = config["classification"].get("val_split", 0.1)
train_transform, val_transform = get_classification_transforms(
img_size, aug
)
full_train = ClassificationDataset(
os.path.join(data_dir, "Training"),
class_names,
img_size,
train_transform,
"train",
)
test_dataset = ClassificationDataset(
os.path.join(data_dir, "Testing"),
class_names,
img_size,
val_transform,
"test",
)
indices = list(range(len(full_train)))
labels = [label for _, label in full_train.samples]
train_indices, val_indices = train_test_split(
indices,
test_size=val_split,
random_state=config["seed"],
stratify=labels,
)
train_dataset = Subset(full_train, train_indices)
val_dataset = ClassificationDataset(
os.path.join(data_dir, "Training"),
class_names,
img_size,
val_transform,
"val",
)
val_dataset.samples = [full_train.samples[i] for i in val_indices]
train_loader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=num_workers,
pin_memory=True,
)
val_loader = DataLoader(
val_dataset,
batch_size=batch_size,
shuffle=False,
num_workers=num_workers,
pin_memory=True,
)
test_loader = DataLoader(
test_dataset,
batch_size=batch_size,
shuffle=False,
num_workers=num_workers,
pin_memory=True,
)
return train_loader, val_loader, test_loader, class_names
def get_segmentation_loaders(
config: dict, num_workers: int = 4
) -> Tuple[DataLoader, DataLoader]:
"""Create train/validation data loaders for segmentation.
Args:
config: Loaded configuration dictionary.
num_workers: Number of workers for data loading.
Returns:
Tuple of (train_loader, val_loader).
"""
set_seed(config["seed"])
data_dir = config["paths"]["data_segmentation"]
img_size = config["segmentation"]["img_size"]
batch_size = config["segmentation"]["batch_size"]
mask_suffix = config["segmentation"].get("mask_suffix", "_mask")
val_split = config["segmentation"].get("val_split", 0.2)
train_transform, val_transform = get_segmentation_transforms(img_size)
full_dataset = SegmentationDataset(
data_dir, mask_suffix=mask_suffix, img_size=img_size, transform=train_transform
)
indices = list(range(len(full_dataset)))
train_indices, val_indices = train_test_split(
indices, test_size=val_split, random_state=config["seed"]
)
train_dataset = Subset(full_dataset, train_indices)
val_dataset = SegmentationDataset(
data_dir,
mask_suffix=mask_suffix,
img_size=img_size,
transform=val_transform,
)
val_dataset.pairs = [full_dataset.pairs[i] for i in val_indices]
train_loader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=num_workers,
pin_memory=True,
)
val_loader = DataLoader(
val_dataset,
batch_size=batch_size,
shuffle=False,
num_workers=num_workers,
pin_memory=True,
)
return train_loader, val_loader
|