数据集:

frgfm/imagewoof

语言:

en

大小:

1K<n<10K

语言创建人:

crowdsourced

批注创建人:

crowdsourced

源数据集:

extended

许可:

apache-2.0
英文

Imagewoof 数据集卡片

数据集概述

这是一个由 Imagewoof 数据集的 10 个类别组成的较小子集,这些类别都是狗的品种,不容易分类。该数据集由 Jeremy Howard 创建,该存储库仅用于在此平台上分享他的工作。库的所有者对数据集的创建、整理或打包不承担任何形式的责任。

支持的任务和排行榜

  • 图像分类:可以使用该数据集训练图像分类模型。

语言

数据集中的类别标签为英文。

数据集结构

数据实例

数据点由图像的URL和其分类标签组成。

{
  'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=320x320 at 0x19FA12186D8>,
  'label': 'Beagle',
}

数据字段

  • 图像:包含图像的 PIL.Image.Image 对象。
  • 标签:图像的预期类别标签。

数据划分

train validation
imagewoof 9025 3929

数据集创建

理由说明

参见 https://huggingface.co/datasets/imagenet-1k#curation-rationale

源数据

初始数据收集和标准化

Imagewoof 是 ImageNet 的一个子集。有关源数据收集的信息可以在 here 中找到。

注释

注释过程

参见 https://huggingface.co/datasets/imagenet-1k#annotation-process

注释者是谁?

参见 https://huggingface.co/datasets/imagenet-1k#who-are-the-annotators

个人和敏感信息

参见 https://huggingface.co/datasets/imagenet-1k#personal-and-sensitive-information

使用数据时的注意事项

数据的社会影响

参见 https://huggingface.co/datasets/imagenet-1k#social-impact-of-dataset

讨论偏见

参见 https://huggingface.co/datasets/imagenet-1k#discussion-of-biases

其他已知限制

参见 https://huggingface.co/datasets/imagenet-1k#other-known-limitations

附加信息

数据集维护者

参见 https://huggingface.co/datasets/imagenet-1k#dataset-curators 和 Jeremy Howard

许可信息

参见 Apache License 2.0

引用信息

@software{Howard_Imagewoof_2019,
    title={Imagewoof: a subset of 10 classes from Imagenet that aren't so easy to classify},
    author={Jeremy Howard},
    year={2019},
    month={March},
    publisher = {GitHub},
    url = {https://github.com/fastai/imagenette#imagewoof}
}

贡献

该数据集由 Jeremy Howard 创建,并于 Github 发布。之后由 @frgfm 集成到 HuggingFace Datasets 中。