ImageDataBunch

Fastai library works with text, tabular data, collaborative filtering (collab) and vision out of the box.

In fact these are the main fastai divisions or modules.

The vision module of the fastai library contains all the necessary functions to define a Dataset and train a model for computer vision tasks.

It contains four different submodules to reach that goal:

• vision.image contains the basic definition of an Image object
• vision.transform contains all the transforms we can use for data augmentation
• vision.data contains the definition of ImageDataBunch and DataBunch
• vision.learner lets you build models

If you use ImageDataBunch class you can put anything that represents an image into it either for: test, train, or validation.

To create ImageDataBunch you can use several methods:

• create_from_ll (from labeled lists)
• from_csv (from a csv file)
• from_df (from DataFrame)
• from_folder (from imagenet style dataset in path with train,valid and test subfolders)
• from_lists (from list of fnames in path)
• from_name_func (from list of fnames in path with label_func)
• from_name_re (uses regular expression to extract the names)

For example, to load MNIST from_folder use this:

from fastai import *
from fastai.vision import *
path = untar_data(URLs.MNIST_SAMPLE)
data = ImageDataBunch.from_folder(path)

To load Oxford PETS dataset you can use this code:

from fastai import *
from fastai.vision import *
path = untar_data(URLs.PETS);
path_anno = path/'annotations'
path_img = path/'images'
pat = r'/([^/]+)_\d+.jpg$'
data = ImageDataBunch.from_name_re(path_img, fnames, pat, ds_tfms=get_transforms(), size=224, bs=bs)

We can print out the datasets:

print(data.valid_ds)
print("...")
print(data.train_ds)
print("...")
print(data.test_ds)

And this will output:

LabelList (1478 items)
x: ImageList
Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224)
y: CategoryList
chihuahua,Maine_Coon,samoyed,pomeranian,japanese_chin
Path: /root/.fastai/data/oxford-iiit-pet/images
...
LabelList (5912 items)
x: ImageList
Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224),Image (3, 224, 224)
y: CategoryList
pomeranian,Maine_Coon,english_cocker_spaniel,Bengal,Sphynx
Path: /root/.fastai/data/oxford-iiit-pet/images
...
None

Notes:

_{Note #1: DataBunch has train_dl, valid_dl and test_dl inside.}

_{Note #2: Use the configuration Config paths.}

Config.DEFAULT_CONFIG = {
        'data_path': '/content/emp/',
        'data_archive_path': '/content/emp/',
        'model_path': '/content/emp/'
}
Config.create('/tmp/myconfig.yml')
Config.DEFAULT_CONFIG_PATH = '/tmp/myconfig.yml'

data_path = Config.data_path()
print(data_path)