TIGER-Lab/Entity-Imagen

---
dataset_info:
- config_name: bird
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 1153768
num_examples: 50
download_size: 1077545
dataset_size: 1153768
- config_name: car
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 5285261
num_examples: 50
download_size: 4433750
dataset_size: 5285261
- config_name: dish
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 24535127
num_examples: 50
download_size: 14104477
dataset_size: 24535127
- config_name: dog
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 1144143
num_examples: 50
download_size: 918873
dataset_size: 1144143
- config_name: landmark
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 305009678
num_examples: 50
download_size: 258293099
dataset_size: 305009678
- config_name: starwar
features:
- name: image
dtype: image
- name: entity_description
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 14444301
num_examples: 50
download_size: 9059561
dataset_size: 14444301
configs:
- config_name: bird
data_files:
- split: train
path: bird/train-*
- config_name: car
data_files:
- split: train
path: car/train-*
- config_name: dish
data_files:
- split: train
path: dish/train-*
- config_name: dog
data_files:
- split: train
path: dog/train-*
- config_name: landmark
data_files:
- split: train
path: landmark/train-*
- config_name: starwar
data_files:
- split: train
path: starwar/train-*
license: mit
task_categories:
- text-to-image
language:
- en
size_categories:
- n<1K
---

This dataset is a replication evaluation set from Re-Imagen (https://arxiv.org/pdf/2209.14491.pdf). The dataset is aimed for evaluating text-to-image generation on rare visual entities.

原始信息汇总

数据集概述

数据集配置

鸟类 (bird)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 1153768
    • 样本数: 50
• 下载大小: 1077545
• 数据集大小: 1153768

汽车 (car)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 5285261
    • 样本数: 50
• 下载大小: 4433750
• 数据集大小: 5285261

菜肴 (dish)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 24535127
    • 样本数: 50
• 下载大小: 14104477
• 数据集大小: 24535127

狗 (dog)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 1144143
    • 样本数: 50
• 下载大小: 918873
• 数据集大小: 1144143

地标 (landmark)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 305009678
    • 样本数: 50
• 下载大小: 258293099
• 数据集大小: 305009678

星球大战 (starwar)

• 特征:
  • image: 图像
  • entity_description: 字符串
  • prompt: 字符串
• 分割:
  • train:
    • 字节数: 14444301
    • 样本数: 50
• 下载大小: 9059561
• 数据集大小: 14444301

数据文件配置

• 鸟类 (bird):
  • train: bird/train-*
• 汽车 (car):
  • train: car/train-*
• 菜肴 (dish):
  • train: dish/train-*
• 狗 (dog):
  • train: dog/train-*
• 地标 (landmark):
  • train: landmark/train-*
• 星球大战 (starwar):
  • train: starwar/train-*

其他信息

• 许可证: MIT
• 任务类别: 文本到图像
• 语言: 英语
• 大小类别: n<1K