floschne/xgqa_1k

---
dataset_info:
features:
- name: question
dtype: string
- name: answer
dtype: string
- name: full_answer
dtype: string
- name: image_id
dtype: string
- name: image
struct:
- name: bytes
dtype: binary
- name: path
dtype: 'null'
splits:
- name: bn
num_bytes: 51624194
num_examples: 1000
- name: de
num_bytes: 51582232
num_examples: 1000
- name: en
num_bytes: 51579211
num_examples: 1000
- name: id
num_bytes: 51590256
num_examples: 1000
- name: ko
num_bytes: 51587731
num_examples: 1000
- name: pt
num_bytes: 51579268
num_examples: 1000
- name: ru
num_bytes: 51602287
num_examples: 1000
- name: zh
num_bytes: 51572077
num_examples: 1000
download_size: 412467532
dataset_size: 412717256
configs:
- config_name: default
data_files:
- split: bn
path: data/bn-*
- split: de
path: data/de-*
- split: en
path: data/en-*
- split: id
path: data/id-*
- split: ko
path: data/ko-*
- split: pt
path: data/pt-*
- split: ru
path: data/ru-*
- split: zh
path: data/zh-*
license: cc-by-4.0
task_categories:
- visual-question-answering
language:
- bn
- de
- en
- id
- ko
- pt
- ru
- zh
pretty_name: xGQA
size_categories:
- 1K<n<10K
---

# xGQA 1K

### This is a 1K subset of the `few_shot-test` split of the xGQA dataset

Please find the original repository here: https://github.com/adapter-hub/xGQA

If you use this dataset, please cite the original authors:

```bibtex
@inproceedings{pfeiffer-etal-2021-xGQA,
title={{xGQA: Cross-Lingual Visual Question Answering}},
author={ Jonas Pfeiffer and Gregor Geigle and Aishwarya Kamath and Jan-Martin O. Steitz and Stefan Roth and Ivan Vuli{\'{c}} and Iryna Gurevych},
booktitle = "Findings of the Association for Computational Linguistics: ACL 2022",
month = May,
year = "2022",
url = "https://arxiv.org/pdf/2109.06082.pdf",
publisher = "Association for Computational Linguistics",
}
```

This subset was sampled so that all languages contain the same images and questions based on the `imageId` and `semanticStr` in the original dataset. In other words, this subset is still parallel.

### How to read the image
Due to a [bug](https://github.com/huggingface/datasets/issues/4796), the images cannot be stored as PIL.Image.Images directly but need to be converted to dataset.Images-. Hence, to load them, this additional step is required:

```python
from datasets import Image, load_dataset

ds = load_dataset("floschne/xgqa_1k", split="en")
ds.map(
lambda sample: {
"image_t": [Image().decode_example(img) for img in sample["image"]],
},
remove_columns=["image"],
).rename_columns({"image_t": "image"})
```

原始信息汇总

数据集概述

数据集名称

• xGQA 1K

数据集特征

• question: 数据类型为字符串
• answer: 数据类型为字符串
• full_answer: 数据类型为字符串
• image_id: 数据类型为字符串
• image: 结构包括
  • bytes: 数据类型为二进制
  • path: 数据类型为空

数据集分割

• bn: 1000个样本，总字节数51624194
• de: 1000个样本，总字节数51582232
• en: 1000个样本，总字节数51579211
• id: 1000个样本，总字节数51590256
• ko: 1000个样本，总字节数51587731
• pt: 1000个样本，总字节数51579268
• ru: 1000个样本，总字节数51602287
• zh: 1000个样本，总字节数51572077

数据集大小

• 下载大小: 412467532字节
• 数据集大小: 412717256字节

配置

• config_name: default
• data_files:
  • split: 不同语言的分割
  • path: 对应语言数据的路径格式

许可

• cc-by-4.0

任务类别

• visual-question-answering

语言

• bn, de, en, id, ko, pt, ru, zh

大小类别

• 1K<n<10K