Virtue-AI-HUB/image-moderation

该数据集主要用于图像内容分类，包含图像、标签和来源三个特征。标签特征详细分类了16种不同的内容类型，如仇恨、暴力、性内容等。数据集分为训练集、测试集和离群数据集，分别包含27889、3084和4997个样本。

This dataset is primarily used for image content classification, featuring images, labels, and sources. The label feature categorizes 16 different types of content, such as hate, violence, sexual content, etc. The dataset is divided into training, testing, and out-of-distribution sets, containing 27889, 3084, and 4997 samples respectively.

原始信息汇总

数据集概述

特征

• image: 图像数据
• label: 标签序列，包含以下类别：
  • 0: NA: None applying / Safe
  • 1: O1: Hate, Humiliation, Harassment
  • 2: O2: Violence, Harm, or Cruelty
  • 3: O3: Sexual Content
  • 4: O4: Nudity Content
  • 5: O5: Criminal Planning
  • 6: O6: Weapons or Substance Abuse
  • 7: O7: Self-Harm
  • 8: O8: Animal Cruelty
  • 9: O9: Disasters or Emergencies
  • 10: 10: Political Content
  • 11: Shocking
  • 12: Public and personal health
  • 13: Spam
  • 14: Insulting Gesture
  • 15: Blood
• source: 字符串类型，表示数据来源

数据集划分

• train: 训练集，包含27889个样本，大小为259861535.802字节
• test: 测试集，包含3084个样本，大小为29236917.54字节
• ood: 未知分布数据集，包含4997个样本，大小为44402490.192字节

数据集大小

• 下载大小: 330978039字节
• 数据集总大小: 333500943.53400004字节

配置

• config_name: default
  • data_files:
    • train: data/train-*
    • test: data/test-*
    • ood: data/ood-*