| import os |
| import torch |
| import torch.utils.data as data |
| import numpy as np |
| from PIL import Image |
| import logging |
| from .data_path import get_dataset_path |
|
|
|
|
| class NIH(data.Dataset): |
| pathologies = [ 'Atelectasis', 'Cardiomegaly', 'Effusion', 'Infiltration', 'Mass', 'Nodule', 'Pneumonia', |
| 'Pneumothorax', 'Consolidation', 'Edema', 'Emphysema', 'Fibrosis', 'Pleural_Thickening', 'Hernia'] |
|
|
| def __init__(self, split="train", data_pct=1.0, transform=None, return_label=False): |
| super().__init__() |
| self.root = get_dataset_path('NIH') |
| self.return_label = return_label |
| self.num_classes = len(self.pathologies) |
| self.split = split |
| self.transform = transform |
| |
| |
| |
| |
| |
| |
| self.nih_proc = False |
| if not return_label: |
| logging.info('NIH for Pretrain on train_val') |
| with open(os.path.join(self.root, 'train_val_list.txt'), 'r') as f: |
| lines = f.readlines() |
| if self.nih_proc: |
| self.data = [os.path.join(self.root, 'NIH_proc', l.strip()) for l in lines if l.strip() != ''] |
| else: |
| self.data = [os.path.join(self.root, 'NIH_512_jpg', l.strip()) for l in lines if l.strip() != ''] |
| logging.info(f'NIH Num Pretrain: {len(self.data)}') |
| else: |
| assert split in ('train', 'val', 'test') |
| self.data = [] |
| self.img_label = [] |
| img_root = os.path.join(self.root, 'NIH_512') |
| if not os.path.exists(img_root): |
| img_root = img_root.replace('NIH_512', 'NIH_512_jpg') |
| with open(f'annotations/nih/Xray14_{split}_official.txt', "r") as fileDescriptor: |
| line = True |
| while line: |
| line = fileDescriptor.readline() |
| if line: |
| lineItems = line.split() |
| imagePath = os.path.join(img_root, lineItems[0]) |
|
|
| imageLabel = lineItems[1:self.num_classes + 1] |
| imageLabel = [int(i) for i in imageLabel] |
|
|
| self.data.append(imagePath) |
| self.img_label.append(imageLabel) |
| if data_pct < 1.0 and split == 'train': |
| state = np.random.RandomState(seed=42) |
| num_data = len(self.data) |
| indices = state.choice(np.arange(num_data), size=int(data_pct * num_data), replace=False) |
| self.data = [self.data[idx] for idx in indices] |
| self.img_label = [self.img_label[idx] for idx in indices] |
| logging.info(f'NIH Downstream {split} set: {len(self.data)}') |
|
|
| def __getitem__(self, index): |
| if not self.return_label: |
| img_path = self.data[index] |
| else: |
| img_path = self.data[index] |
| label = self.img_label[index] |
| |
| |
| |
| |
| |
| if 'jpg' in img_path: |
| img = Image.open(img_path.replace('.png', '.jpg')).convert('RGB') |
| else: |
| img = Image.open(img_path).convert('RGB') |
| if self.transform != None: |
| img = self.transform(img) |
| |
| if not self.return_label: |
| return img |
| else: |
| return img, torch.FloatTensor(label) |
|
|
| def __len__(self): |
| return len(self.data) |
|
|
