| import os |
| import random |
| import sys |
|
|
| import numpy as np |
| import PIL.Image |
| import torch |
| import torch.nn.functional as F |
| from PIL import Image |
| from torch.utils.data import DataLoader, Dataset |
| from torchvision import transforms |
|
|
|
|
| class GenerateCGMHDataset(Dataset): |
| def __init__(self, root_path, transform=None): |
| self.root_path = root_path |
| self.image_path = os.path.join(self.root_path, "Image/") |
| self.label_path = os.path.join(self.root_path, "Label/") |
| self.path_set = [] |
| for path in os.listdir(self.image_path): |
| if path.endswith(".png"): |
| self.path_set.append(os.path.join(self.image_path,path)) |
| if transform == None: |
| self.transform = transforms.Compose([transforms.ToTensor(), transforms.Resize((256, 256))]) |
|
|
| def __len__(self): |
| return len(self.path_set) |
|
|
| def __getitem__(self, item): |
| path = self.path_set[item] |
| image_path = path |
| label_path = path.replace("Image/", "Label/") |
| image = PIL.Image.open(image_path).convert("L") |
| label = PIL.Image.open(label_path).convert("L") |
| image = self.transform(image).float() |
| label = (self.transform(label) > 0.5).float() |
| if random.random() > 0.5: |
| return label, 1, label |
| else: |
| return image, 0, label |
|
|
| class CGMHDataset(Dataset): |
| def __init__(self, root_path, transform=None,if_val = False): |
| self.root_path = root_path |
| self.image_path = os.path.join(self.root_path, "Image/") |
| self.label_path = os.path.join(self.root_path, "Label/") |
| self.path_set = [] |
| for path in os.listdir(self.image_path): |
| if path.endswith(".png"): |
| self.path_set.append(os.path.join(self.image_path,path)) |
| if transform == None: |
| self.transform = transforms.Compose([transforms.ToTensor(), transforms.Resize((256, 256))]) |
|
|
| from utils.stnaugment import STNAugment |
| self.data_aug = STNAugment() |
| self.if_val = if_val |
|
|
| def apply_transforms(self, image,label, transform, seed=None): |
| if seed is None: |
| MAX_RAND_VAL = 2147483647 |
| seed = np.random.randint(MAX_RAND_VAL) |
|
|
| if transform is not None: |
| random.seed(seed) |
| torch.random.manual_seed(seed) |
| turn_list = [] |
| turn_list.append(image) |
| turn_list.append(label) |
| turn_list = self.data_aug(turn_list) |
| return turn_list[0],turn_list[1] |
|
|
| def __len__(self): |
| return len(self.path_set) |
|
|
| def __getitem__(self, item): |
| path = self.path_set[item] |
| image_path = path |
| label_path = path.replace("Image/", "Label/") |
| image = PIL.Image.open(image_path).convert("L") |
| label = PIL.Image.open(label_path).convert("L") |
| image = self.transform(image).float() |
| label = (self.transform(label) > 0.5).float() |
| image,label = self.apply_transforms(image,label,transforms) |
| if self.if_val: |
| return image,label |
| else: |
| if_label = random.random() > 0.5 |
| if if_label: |
| return (label) * 2 - 1, 1, label |
| else: |
| return (image) * 2 - 1, 0, label |
|
|
|
|
| def split_train_and_val(dataset,split_ratio = 0.9): |
| from sklearn.model_selection import StratifiedShuffleSplit |
| labels = [0 for i in range(len(dataset))] |
| ss = StratifiedShuffleSplit(n_splits=1, test_size=1 - split_ratio, random_state=0) |
| train_indices, valid_indices = list(ss.split(np.array(labels)[:, np.newaxis], labels))[0] |
| dst_train = torch.utils.data.Subset(dataset, train_indices) |
| dst_test = torch.utils.data.Subset(dataset, valid_indices) |
| return dst_train,dst_test |
|
|
