yujiepan's picture
Update README.md
a4e7623 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: label
      dtype: int64
  splits:
    - name: train
      num_bytes: 723660480
      num_examples: 2048
  download_size: 723705789
  dataset_size: 723660480
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Dataset Card for "imagenet-1k-train-sampled2048"

import numpy as np
from datasets import Dataset
import timm.data


def get_dataset():
    dataset = timm.data.create_dataset(
        root='[LOCAL IMAGENET FOLDER]',
        name='',
        split='train',
        is_training=False,
    )
    sampled_indices = np.random.default_rng(42).choice(len(dataset), size=2048, replace=False)
    images = [dataset[i][0] for i in sampled_indices]
    labels = [dataset[i][1] for i in sampled_indices]
    return images, labels


images, labels = get_dataset()
dataset = Dataset.from_dict({"image": images, "label": labels})
dataset.push_to_hub('yujiepan/imagenet-1k-train-sampled2048')