from PIL import Image
from torchvision.datasets import folder
from collections import defaultdict
import csv
import glob
import os
import os.path
import torch.utils.data as data
[docs]class CsvImageFolder(data.Dataset):
"""
A Dataset that reads images from a folder and labels from a csv file.
Arguments:
root (string): directory where the images can be found.
csv_path (string): the path to the csv file containing image filenames and labels.
filename_key (string, optional): the column header of the csv for the column that contains image filenames
(without extensions).
labels_key (string, optional): the column header of the csv for the column that contains labels for each image.
csv_delimiter (string, optional): the character(s) used to separate fields in the csv file.
loader (callable, optional): a function to load a sample given its path.
extensions (list[string], optinoal): a list of allowed extensions. E.g, ``['.jpg', '.tif']``
transform (callable, optional): A function/transform that takes in a sample and returns a transformed version.
E.g, ``transforms.RandomCrop`` for images.
target_transform (callable, optional): A function/transform that takes in the target and transforms it.
"""
def __init__(self, root, csv_path,
filename_key='image_name',
labels_key='tags',
csv_delimiter=',',
labels_delimiter=' ',
loader=folder.default_loader,
extensions=folder.IMG_EXTENSIONS,
transform=None,
target_transform=None):
self.root = root
self.csv_path = csv_path
self.filename_key = filename_key
self.labels_key = labels_key
self.csv_delimiter = csv_delimiter
self.labels_delimiter = labels_delimiter
self.loader = loader
self.extensions = extensions
self.transform = transform
self.target_transform = target_transform
self._found_extensions = None
self._load_found_extensions()
self._load()
def _load(self):
self._rows = []
all_labels = set()
with open(self.csv_path) as csv_file:
reader = csv.DictReader(csv_file, delimiter=self.csv_delimiter)
for row in reader:
self._rows.append(row)
labels = self._get_labels(row)
all_labels.update(labels)
self._labels = sorted(all_labels)
self._label_idxs = {label:i for i, label in enumerate(self._labels)}
def _get_path(self, base_filename):
found_extensions = self.found_extensions
if base_filename in found_extensions:
found_extensions = found_extensions[base_filename]
for extension in self.extensions:
if extension in found_extensions:
return os.path.join(self.root, '%s%s' % (base_filename, extension))
raise RuntimeError(f"No image file with base filename ({base_filename}) "
f"found in folder ({self.root}), valid extensions are: {self.extensions}")
@property
def found_extensions(self):
if self._found_extensions is None:
self._found_extensions = self._load_found_extensions()
return self._found_extensions
def _load_found_extensions(self):
found_extensions = defaultdict(list)
for filename in os.listdir(self.root):
base, ext = os.path.splitext(filename)
found_extensions[base].append(ext)
return dict(found_extensions)
def _get_labels(self, row):
return set(row[self.labels_key].split(self.labels_delimiter))
def __getitem__(self, index):
"""
Args:
index (int): Index
Returns:
tuple: (sample, target) where sample is the image, and target is an array of indices of the target
class.
"""
row = self._rows[index]
base_filename = row[self.filename_key]
path = self._get_path(base_filename)
sample = self.loader(path)
labels = self._get_labels(row)
target = [self._label_idxs[label] for label in labels]
if self.transform is not None:
sample = self.transform(sample)
if self.target_transform is not None:
target = self.target_transform(target)
return sample, target
def __len__(self):
return len(self._rows)
def __repr__(self):
fmt_str = 'Dataset ' + self.__class__.__name__ + '\n'
fmt_str += ' Number of images: {}\n'.format(self.__len__())
fmt_str += ' Root Location: {}\n'.format(self.root)
fmt_str += ' Number of tags: {}\n'.format(len(self._labels))
tmp = ' Transforms (if any): '
fmt_str += '{0}{1}\n'.format(tmp, self.transform.__repr__().replace('\n', '\n' + ' ' * len(tmp)))
tmp = ' Target Transforms (if any): '
fmt_str += '{0}{1}'.format(tmp, self.target_transform.__repr__().replace('\n', '\n' + ' ' * len(tmp)))
return fmt_str