ImageNet in PyTorch

Buy Me a Coffee☕

*My post explains ImageNet.

ImageNet() can use ImageNet dataset as shown below:

*Memos:

• The 1st argument is root(Required-Type:str or pathlib.Path). *An absolute or relative path is possible.
• The 2nd argument is split(Optional-Default:"train"-Type:str): *Memos:
  • "train"(1,281,167 images) or "val"(50,000 images) can be set to it.
  • "test"(100,000 images) isn't supported so I requested the feature on GitHub.
• There is transform argument(Optional-Default:None-Type:callable). *transform= must be used.
• There is target_transform argument(Optional-Default:None-Type:callable). - There is transform argument(Optional-Default:None-Type:callable). *target_transform= must be used.
• There is loader argument(Optional-Default:torchvision.datasets.folder.default_loader-Type:callable). *loader= must be used.
• You have to manually download the dataset(ILSVRC2012_devkit_t12.tar.gz, ILSVRC2012_img_train.tar and ILSVRC2012_img_val.tar to data/, then running ImageNet() extracts and loads the dataset.
• About the label from the classes for the train and validation image indices respectively, tench&Tinca tinca(0) are 0~1299 and 0~49, goldfish&Carassius auratus(1) are 1300~2599 and 50~99, great white shark&white shark&man-eater&man-eating shark&Carcharodon carcharias(2) are 2600~3899 and 100~149, tiger shark&Galeocerdo cuvieri(3) are 3900~5199 and 150~199, hammerhead&hammerhead shark(4) are 5200~6499 and 200~249, electric ray&crampfish&numbfish&torpedo(5) are 6500~7799 and 250~299, stingray(6) is 7800~9099 and 250~299, cock(7) is 9100~10399 and 300~349, hen(8) is 10400~11699 and 350~399, ostrich&Struthio camelus(9) are 11700~12999 and 400~449, etc.

from torchvision.datasets import ImageNet
from torchvision.datasets.folder import default_loader

train_data = ImageNet(
    root="data"
)

train_data = ImageNet(
    root="data",
    split="train",
    transform=None,
    target_transform=None,
    loader=default_loader
)

val_data = ImageNet(
    root="data",
    split="val"
)

len(train_data), len(val_data)
# (1281167, 50000)

train_data
# Dataset ImageNet
#     Number of datapoints: 1281167
#     Root location: D:/data
#     Split: train

train_data.root
# 'data'

train_data.split
# 'train'

print(train_data.transform)
# None

print(train_data.target_transform)
# None

train_data.loader
# <function torchvision.datasets.folder.default_loader(path: str) -> Any>

len(train_data.classes), train_data.classes
# (1000,
#  [('tench', 'Tinca tinca'), ('goldfish', 'Carassius auratus'),
#   ('great white shark', 'white shark', 'man-eater', 'man-eating shark',
#    'Carcharodon carcharias'), ('tiger shark', 'Galeocerdo cuvieri'),
#   ('hammerhead', 'hammerhead shark'), ('electric ray', 'crampfish',
#    'numbfish', 'torpedo'), ('stingray',), ('cock',), ('hen',),
#   ('ostrich', 'Struthio camelus'), ..., ('bolete',), ('ear', 'spike',
#    'capitulum'), ('toilet tissue', 'toilet paper', 'bathroom tissue')])

train_data[0]
# (<PIL.Image.Image image mode=RGB size=250x250>, 0)

train_data[1]
# (<PIL.Image.Image image mode=RGB size=200x150>, 0)

train_data[2]
# (<PIL.Image.Image image mode=RGB size=500x375>, 0)

train_data[1300]
# (<PIL.Image.Image image mode=RGB size=640x480>, 1)

train_data[2600]
# (<PIL.Image.Image image mode=RGB size=500x375>, 2)

val_data[0]
# (<PIL.Image.Image image mode=RGB size=500x375>, 0)

val_data[1]
# (<PIL.Image.Image image mode=RGB size=500x375>, 0)

val_data[2]
# (<PIL.Image.Image image mode=RGB size=500x375>, 0)

val_data[50]
# (<PIL.Image.Image image mode=RGB size=500x500>, 1)

val_data[100]
# (<PIL.Image.Image image mode=RGB size=679x444>, 2)

import matplotlib.pyplot as plt

def show_images(data, ims, main_title=None):
    plt.figure(figsize=[12, 6])
    plt.suptitle(t=main_title, y=1.0, fontsize=14)
    for i, j in enumerate(iterable=ims, start=1):
        plt.subplot(2, 5, i)
        im, lab = data[j]
        plt.imshow(X=im)
        plt.title(label=lab)
    plt.tight_layout(h_pad=3.0)
    plt.show()

train_ims = [0, 1, 2, 1300, 2600, 3900, 5200, 6500, 7800, 9100]
val_ims = [0, 1, 2, 50, 100, 150, 200, 250, 300, 350]

show_images(data=train_data, ims=train_ims, main_title="train_data")
show_images(data=val_data, ims=val_ims, main_title="val_data")