Datasets:

leekwoon
/

oxfordpet_dataset_backup

# huggingface_hub 설치 (필요시)
pip install huggingface_hub

# 다운로드
huggingface-cli download leekwoon/oxfordpet_dataset_backup --repo-type dataset --local-dir ./oxfordpet_data

# 무결성 확인 (선택사항)
cd oxfordpet_data
md5sum -c checksums.md5

# 파일 합치기 및 압축 해제
cat data.tar.gz.part_* | tar -xzvf -

Dataset Information

Overview

The Oxford-IIIT Pet Dataset is a 37-category pet dataset with roughly 200 images for each class. The dataset contains images of cats and dogs belonging to 37 breeds, with breed annotations and segmentation masks. It's widely used for fine-grained classification, object detection, and segmentation tasks.

Statistics

Total Images: ~7,390
Training Images: ~3,680
Validation Images: ~370
Test Images: ~3,340
Number of Classes: 37 (12 cat breeds + 25 dog breeds)
Image Format: JPEG
Average Resolution: Variable (approximately 400x500 pixels)

Directory Structure

oxfordpet/
├── train/                  # Training images
│   ├── Abyssinian/        # Cat breed
│   ├── Bengal/            # Cat breed
│   ├── Birman/            # Cat breed
│   ├── Bombay/            # Cat breed
│   ├── British_Shorthair/
│   ├── Egyptian_Mau/
│   ├── Maine_Coon/
│   ├── Persian/
│   ├── Ragdoll/
│   ├── Russian_Blue/
│   ├── Siamese/
│   ├── Sphynx/
│   ├── american_bulldog/  # Dog breed
│   ├── american_pit_bull_terrier/
│   ├── basset_hound/
│   ├── beagle/
│   ├── boxer/
│   ├── chihuahua/
│   ├── english_cocker_spaniel/
│   ├── english_setter/
│   ├── german_shorthaired/
│   ├── great_pyrenees/
│   ├── havanese/
│   ├── japanese_chin/
│   ├── keeshond/
│   ├── leonberger/
│   ├── miniature_pinscher/
│   ├── newfoundland/
│   ├── pomeranian/
│   ├── pug/
│   ├── saint_bernard/
│   ├── samoyed/
│   ├── scottish_terrier/
│   ├── shiba_inu/
│   ├── staffordshire_bull_terrier/
│   └── yorkshire_terrier/
├── val/                    # Validation images
│   └── [same structure as train]
└── test/                   # Test images
    └── [same structure as train]

Pet Breeds

Cat Breeds (12)

Abyssinian
Bengal
Birman
Bombay
British Shorthair
Egyptian Mau
Maine Coon
Persian
Ragdoll
Russian Blue
Siamese
Sphynx

Dog Breeds (25)

American Bulldog
American Pit Bull Terrier
Basset Hound
Beagle
Boxer
Chihuahua
English Cocker Spaniel
English Setter
German Shorthaired Pointer
Great Pyrenees
Havanese
Japanese Chin
Keeshond
Leonberger
Miniature Pinscher
Newfoundland
Pomeranian
Pug
Saint Bernard
Samoyed
Scottish Terrier
Shiba Inu
Staffordshire Bull Terrier
Wheaten Terrier
Yorkshire Terrier

Usage

Loading with PyTorch

from torchvision import datasets, transforms

transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                       std=[0.229, 0.224, 0.225]),
])

# Load training data
train_dataset = datasets.ImageFolder('oxfordpet/train', transform=transform)

# Load validation data
val_dataset = datasets.ImageFolder('oxfordpet/val', transform=transform)

# Load test data
test_dataset = datasets.ImageFolder('oxfordpet/test', transform=transform)

# Get class names
class_names = train_dataset.classes

Loading with TensorFlow

import tensorflow as tf

# Define preprocessing
def preprocess_image(image, label):
    image = tf.image.resize(image, [224, 224])
    image = tf.cast(image, tf.float32) / 255.0
    return image, label

# Load dataset
train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    'oxfordpet/train',
    image_size=(224, 224),
    batch_size=32
).map(preprocess_image)

val_ds = tf.keras.preprocessing.image_dataset_from_directory(
    'oxfordpet/val',
    image_size=(224, 224),
    batch_size=32
).map(preprocess_image)

Key Challenges

This dataset presents several interesting challenges:

Fine-grained classification: Distinguishing between similar breeds
Pose variation: Pets in various poses and orientations
Scale variation: Different distances from camera
Occlusion: Partial visibility of pets
Background variation: Indoor and outdoor settings
Intra-class variation: Same breed can look quite different

Dataset Features

Breed classification: 37 different breeds of cats and dogs
Species classification: Binary classification (cat vs dog)
Segmentation masks: Available in original dataset (not included in this backup)
Head bounding boxes: Available in original dataset (not included in this backup)

Important Notes

Images have variable sizes and aspect ratios
Some breeds have more images than others
Images are real-world photos with natural backgrounds
File naming convention: {breed}_{number}.jpg
This is a popular benchmark for fine-grained classification

Citation

If you use this dataset, please cite the original paper:

@InProceedings{parkhi12a,
  author = {Parkhi, O. M. and Vedaldi, A. and Zisserman, A. and Jawahar, C.~V.},
  title = {Cats and Dogs},
  booktitle = {IEEE Conference on Computer Vision and Pattern Recognition},
  year = {2012}
}

License

The dataset is available for research purposes. The images were collected from the internet and each image may have its own copyright. For commercial use, please refer to the original dataset documentation.

Acknowledgments

This dataset was created by the Visual Geometry Group at the University of Oxford in collaboration with IIIT Hyderabad.