I am a new student who just learned pytorch. I want to use Deeplab V3+ to train my data set. I downloaded a code from github and used Deepfashion data set，How do I load my own camvid data set?
The data set code is as follows:
from future import print_function, division
import os
from PIL import Image
import json
import numpy as np
from torch.utils.data import Dataset
from torchvision import transforms
from preprocessing import custom_transforms as tr
import random
class DeepFashionSegmentation(Dataset):
“”"
DeepFashion dataset
“”"
NUM_CLASSES = 14
def __init__(self,
config,
base_dir=config[‘dataset’][‘base_path’],
split='train',
):
super().__init__()
self._base_dir = config['dataset']['base_path']
self._image_dir = os.path.join(self._base_dir, 'train', 'image')
self._cat_dir = os.path.join(self._base_dir, 'labels')
self.config = config
self.split = split
with open(os.path.join(self._base_dir, 'train_val_test.json')) as f:
self.full_dataset = json.load(f)
self.images = []
self.categories = []
self.num_classes = self.config['network']['num_classes']
self.shuffle_dataset()
# Display stats
print('Number of images in {}: {:d}'.format(split, len(self.images)))
def shuffle_dataset(self):
#reset lists
self.images.clear()
self.categories.clear()
dataset = self.full_dataset[self.split]
if self.split == 'train' and self.config['training']['train_on_subset']['enabled']:
fraction = self.config['training']['train_on_subset']['dataset_fraction']
sample = int(len(dataset) * fraction)
dataset = random.sample(dataset, sample)
for item in dataset:
self.images.append(os.path.join(self._image_dir, item['image']))#train中的image
self.categories.append(os.path.join(self._cat_dir, item['annotation']))
#be sure that total dataset size is divisible by 2
if len(self.images) % 2 != 0:
self.images.append(os.path.join(self._image_dir, item['image']))
self.categories.append(os.path.join(self._cat_dir, item['annotation']))
assert (len(self.images) == len(self.categories))
print(self.images[0])
print(len(self.images))
def __len__(self):
return len(self.images)
def __getitem__(self, index):
_img, _target = self._make_img_gt_point_pair(index)
sample = {'image': _img, 'label': _target}
#for split in self.split:
if self.split == "train":
print(‘train’)
return self.transform_tr(sample)
elif self.split == 'val':
print(‘val’)
return self.transform_val(sample)
elif self.split == 'test':
print(‘in return’)
return self.transform_val(sample)
def _make_img_gt_point_pair(self, index):
_img = Image.open(self.images[index]).convert('RGB')
_target = Image.open(self.categories[index])
return _img, _target
def transform_tr(self, sample):
composed_transforms = transforms.Compose([
tr.RandomHorizontalFlip(),
tr.RandomScaleCrop(base_size=self.config['image']['base_size'], crop_size=self.config['image']['crop_size']),
tr.RandomGaussianBlur(),
tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
tr.ToTensor()])
return composed_transforms(sample)
def transform_val(self, sample):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=crop_size),
tr.FixScaleCrop(crop_size=self.config['image']['crop_size']),
tr.FixScaleCrop(crop_size=513),
tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
tr.ToTensor()])
return composed_transforms(sample)
@staticmethod
def preprocess(sample, crop_size=513):
composed_transforms = transforms.Compose([
tr.FixScaleCrop(crop_size=crop_size),
tr.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)),
tr.ToTensor()])
return composed_transforms(sample)
def __str__(self):
return 'DeepFashion2(split=' + str(self.split) + ')'
The data loader code is as follows:
from torch.utils.data import DataLoader
from data_generators.deepfashion import DeepFashionSegmentation
def initialize_data_loader(config):
if config['dataset']['dataset_name'] == 'deepfashion':
train_set = DeepFashionSegmentation(config, split='train')
val_set = DeepFashionSegmentation(config, split='val')
test_set = DeepFashionSegmentation(config, split='test')
else:
raise Exception('dataset not implemented yet!')
num_classes = train_set.num_classes
train_loader = DataLoader(train_set, batch_size=config['training']['batch_size'], shuffle=True, num_workers=config['training']['workers'], pin_memory=True)
val_loader = DataLoader(val_set, batch_size=config['training']['batch_size'], shuffle=False, num_workers=config['training']['workers'], pin_memory=True)
test_loader = DataLoader(test_set, batch_size=config['training']['batch_size'], shuffle=False, num_workers=config['training']['workers'], pin_memory=True)
return train_loader, val_loader, test_loader, num_classes
My own Camvid data set format is as follows:
