conceptual_captions

参考:

使用以下命令在 TFDS 中加载此数据集:

ds = tfds.load('huggingface:conceptual_captions')
  • 说明
Image captioning dataset
The resulting dataset (version 1.1) has been split into Training, Validation, and Test splits. The Training split consists of 3,318,333 image-URL/caption pairs, with a total number of 51,201 total token types in the captions (i.e., total vocabulary). The average number of tokens per captions is 10.3 (standard deviation of 4.5), while the median is 9.0 tokens per caption. The Validation split consists of 15,840 image-URL/caption pairs, with similar statistics.
  • 许可:无已知许可
  • 版本:1.0.0
  • 拆分
拆分 样本
'train' 3318333
'validation' 15840
  • 特征
{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "caption": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

unlabeled

使用以下命令在 TFDS 中加载此数据集:

ds = tfds.load('huggingface:conceptual_captions/unlabeled')
  • 说明
Google's Conceptual Captions dataset has more than 3 million images, paired with natural-language captions.
In contrast with the curated style of the MS-COCO images, Conceptual Captions images and their raw descriptions are harvested from the web,
and therefore represent a wider variety of styles. The raw descriptions are harvested from the Alt-text HTML attribute associated with web images.
The authors developed an automatic pipeline that extracts, filters, and transforms candidate image/caption pairs, with the goal of achieving a balance of cleanliness,
informativeness, fluency, and learnability of the resulting captions.
  • 许可:数据集可以免费用于任何目的,但我们将感谢您承认 Google LLC(“Google”)作为数据源。数据集按“原样”提供,不提供任何明示或暗示的保证。对于因使用数据集而导致的任何直接或间接损害,Google 不承担任何责任。

  • 版本:0.0.0

  • 拆分

拆分 样本
'train' 3318333
'validation' 15840
  • 特征
{
    "image_url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "caption": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

labeled

使用以下命令在 TFDS 中加载此数据集:

ds = tfds.load('huggingface:conceptual_captions/labeled')
  • 说明
Google's Conceptual Captions dataset has more than 3 million images, paired with natural-language captions.
In contrast with the curated style of the MS-COCO images, Conceptual Captions images and their raw descriptions are harvested from the web,
and therefore represent a wider variety of styles. The raw descriptions are harvested from the Alt-text HTML attribute associated with web images.
The authors developed an automatic pipeline that extracts, filters, and transforms candidate image/caption pairs, with the goal of achieving a balance of cleanliness,
informativeness, fluency, and learnability of the resulting captions.
  • 许可:数据集可以免费用于任何目的,但我们将感谢您承认 Google LLC(“Google”)作为数据源。数据集按“原样”提供,不提供任何明示或暗示的保证。对于因使用数据集而导致的任何直接或间接损害,Google 不承担任何责任。

  • 版本:0.0.0

  • 拆分

拆分 样本
'train' 2007090
  • 特征
{
    "image_url": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "caption": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "labels": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "MIDs": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "confidence_scores": {
        "feature": {
            "dtype": "float64",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}