DEV Community

Super Kai (Kazuya Ito)
Super Kai (Kazuya Ito)

Posted on • Edited on

OxfordIIITPet in PyTorch

Buy Me a Coffee

*Memos:

OxfordIIITPet() can use Oxford-IIIT Pet 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:"trainval"-Type:str). *"trainval"(3,680 images) or "test"(3,669 images) can be set to it.
  • The 3rd argument is target_types(Optional-Default:"attr"-Type:str or list of str): *Memos:
    • "category", "binary-category" and/or "segmentation" can be set to it: *Memos:
    • "category" is for the label from 37 categories(classes).
    • "binary-category" is for the label of cat(0) or dog(1).
    • "segmentation" is for a segmentation trimap image.
    • An empty tuple or list can also be set to it.
    • The multiple same values can be set to it.
    • If the order of values is different, the order of their elements is also different.
  • The 4th argument is transforms(Optional-Default:None-Type:callable).
  • The 5th argument is transform(Optional-Default:None-Type:callable).
  • The 6th argument is target_transform(Optional-Default:None-Type:callable).
  • The 7th argument is download(Optional-Default:False-Type:bool): *Memos:
    • If it's True, the dataset is downloaded from the internet and extracted(unzipped) to root.
    • If it's True and the dataset is already downloaded, it's extracted.
    • If it's True and the dataset is already downloaded and extracted, nothing happens.
    • It should be False if the dataset is already downloaded and extracted because it's faster.
    • You can manually download and extract the dataset(images.tar.gz and annotations.tar.gz) from here to data/oxford-iiit-pet/.
  • About the label from the categories(classes) for the train image indices, Abyssinian(0) is 0~49, American Bulldog(1) is 50~99, American Pit Bull Terrier(2) is 100~149, Basset Hound(3) is 150~199, Beagle(4) is 200~249, Bengal(5) is 250~299, Birman(6) is 300~349, Bombay(7) is 350~398, Boxer(8) is 399~448, British Shorthair(9) is 449~498, etc.
  • About the label from the categories(classes) for the test image indices, Abyssinian(0) is 0~97, American Bulldog(1) is 98~197, American Pit Bull Terrier(2) is 198~297, Basset Hound(3) is 298~397, Beagle(4) is 398~497, Bengal(5) is 498~597, Birman(6) is 598~697, Bombay(7) is 698~785, Boxer(8) is 786~884, British Shorthair(9) is 885~984, etc.
from torchvision.datasets import OxfordIIITPet

trainval_cate_data = OxfordIIITPet(
    root="data"
)

trainval_cate_data = OxfordIIITPet(
    root="data",
    split="trainval",
    target_types="category",
    transform=None,
    target_transform=None,
    download=False
)

trainval_bincate_data = OxfordIIITPet(
    root="data",
    split="trainval",
    target_types="binary-category"
)

test_seg_data = OxfordIIITPet(
    root="data",
    split="test",
    target_types="segmentation"
)

test_empty_data = OxfordIIITPet(
    root="data",
    split="test",
    target_types=[]
)

test_all_data = OxfordIIITPet(
    root="data",
    split="test",
    target_types=["category", "binary-category", "segmentation"]
)

len(trainval_cate_data), len(trainval_bincate_data)
# (3680, 3680)

len(test_seg_data), len(test_empty_data), len(test_all_data)
# (3669, 3669, 3669)

trainval_cate_data
# Dataset OxfordIIITPet
#     Number of datapoints: 3680
#     Root location: data

trainval_cate_data.root
# 'data'

trainval_cate_data._split
# 'trainval'

trainval_cate_data._target_types
# ['category']

print(trainval_cate_data.transform)
# None

print(trainval_cate_data.target_transform)
# None

trainval_cate_data._download
# <bound method OxfordIIITPet._download of Dataset OxfordIIITPet
#     Number of datapoints: 3680
#     Root location: data>

len(trainval_cate_data.classes), trainval_cate_data.classes
# (37,
#  ['Abyssinian', 'American Bulldog', 'American Pit Bull Terrier',
#   'Basset Hound', 'Beagle', 'Bengal', 'Birman', 'Bombay', 'Boxer',
#   'British Shorthair', ..., 'Wheaten Terrier', 'Yorkshire Terrier'])

trainval_cate_data[0]
# (<PIL.Image.Image image mode=RGB size=394x500>, 0)

trainval_cate_data[1]
# (<PIL.Image.Image image mode=RGB size=450x313>, 0)

trainval_cate_data[2]
# (<PIL.Image.Image image mode=RGB size=500x465>, 0)

trainval_bincate_data[0]
# (<PIL.Image.Image image mode=RGB size=394x500>, 0)

trainval_bincate_data[1]
# (<PIL.Image.Image image mode=RGB size=450x313>, 0)

trainval_bincate_data[2]
# (<PIL.Image.Image image mode=RGB size=500x465>, 0)

test_seg_data[0]
# (<PIL.Image.Image image mode=RGB size=300x225>,
#  <PIL.PngImagePlugin.PngImageFile image mode=L size=300x225>)

test_seg_data[1]
# (<PIL.Image.Image image mode=RGB size=300x225>,
#  <PIL.PngImagePlugin.PngImageFile image mode=L size=300x225>)

test_seg_data[2]
# (<PIL.Image.Image image mode=RGB size=229x300>,
#  <PIL.PngImagePlugin.PngImageFile image mode=L size=229x300>)

test_empty_data[0]
# (<PIL.Image.Image image mode=RGB size=300x225>, None)

test_empty_data[1]
# (<PIL.Image.Image image mode=RGB size=300x225>, None)

test_empty_data[2]
# (<PIL.Image.Image image mode=RGB size=229x300>, None)

test_all_data[0]
# (<PIL.Image.Image image mode=RGB size=300x225>,
#  (0, 0, <PIL.PngImagePlugin.PngImageFile image mode=L size=300x225>))

test_all_data[1]
# (<PIL.Image.Image image mode=RGB size=300x225>,
#  (0, 0, <PIL.PngImagePlugin.PngImageFile image mode=L size=300x225>))

test_all_data[2]
# (<PIL.Image.Image image mode=RGB size=229x300>,
#  (0, 0, <PIL.PngImagePlugin.PngImageFile image mode=L size=229x300>))

import matplotlib.pyplot as plt

def show_images(data, ims, main_title=None):
    if len(data._target_types) == 0:      
        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, _ = data[j]
            plt.imshow(X=im)
    elif len(data._target_types) == 1:
        if data._target_types[0] == "category":
            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, cate = data[j]
                plt.imshow(X=im)
                plt.title(label=cate)
        elif data._target_types[0] == "binary-category":
            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, bincate = data[j]
                plt.imshow(X=im)
                plt.title(label=bincate)
        elif data._target_types[0] == "segmentation":
            plt.figure(figsize=(12, 12))
            plt.suptitle(t=main_title, y=1.0, fontsize=14)
            for i, j in enumerate(iterable=ims, start=1):
                im, seg = data[j]
                if 1 <= i and i <= 5:
                    plt.subplot(4, 5, i)
                    plt.imshow(X=im)
                    plt.subplot(4, 5, i+5)
                    plt.imshow(X=seg)
                if 6 <= i and i <= 10:
                    plt.subplot(4, 5, i+5)
                    plt.imshow(X=im)
                    plt.subplot(4, 5, i+10)
                    plt.imshow(X=seg)
    elif len(data._target_types) == 3:
        plt.figure(figsize=(12, 12))
        plt.suptitle(t=main_title, y=1.0, fontsize=14)
        for i, j in enumerate(iterable=ims, start=1):
            im, (cate, bincate, seg) = data[j]
            if 1 <= i and i <= 5:
                plt.subplot(4, 5, i)
                plt.imshow(X=im)
                plt.title(label=f"{cate}, {bincate}")
                plt.subplot(4, 5, i+5)
                plt.imshow(X=seg)
            if 6 <= i and i <= 10:
                plt.subplot(4, 5, i+5)
                plt.imshow(X=im)
                plt.title(label=f"{cate}, {bincate}")
                plt.subplot(4, 5, i+10)
                plt.imshow(X=seg)
    plt.tight_layout(h_pad=3.0)
    plt.show()

train_ims = (0, 1, 2, 50, 100, 150, 200, 250, 300, 350)
test_ims = (0, 1, 2, 98, 198, 298, 398, 498, 598, 698)

show_images(data=trainval_cate_data, ims=train_ims,
            main_title="trainval_cate_data")
show_images(data=trainval_bincate_data, ims=train_ims, 
            main_title="trainval_bincate_data")
show_images(data=test_seg_data, ims=test_ims,
            main_title="test_seg_data")
show_images(data=test_empty_data, ims=test_ims,
            main_title="test_empty_data")
show_images(data=test_all_data, ims=test_ims,
            main_title="test_all_data")
Enter fullscreen mode Exit fullscreen mode

Image description

Image description

Image description

Image description

Image description

Image of Datadog

The Future of AI, LLMs, and Observability on Google Cloud

Datadog sat down with Google’s Director of AI to discuss the current and future states of AI, ML, and LLMs on Google Cloud. Discover 7 key insights for technical leaders, covering everything from upskilling teams to observability best practices

Learn More

Top comments (0)

Image of Docusign

🛠️ Bring your solution into Docusign. Reach over 1.6M customers.

Docusign is now extensible. Overcome challenges with disconnected products and inaccessible data by bringing your solutions into Docusign and publishing to 1.6M customers in the App Center.

Learn more