dataset_info:
features:
- name: image
dtype: image
- name: name_en
dtype: string
- name: name_zh
dtype: string
- name: text_en
dtype: string
- name: text_zh
dtype: string
splits:
- name: train
num_bytes: 117645424
num_examples: 898
download_size: 117512478
dataset_size: 117645424
Dataset Card for Pokémon wiki captions
This project is inspired by pokmon-blip-captions, where the captions are all generated by pre-trained BLIP without any manual effort. However, the quality and accuracy of their captions are not satisfactory enough, which leaves it known whether better captions lead to better results. This motivates our dataset.
Example
General attribute, looks like a little monkey, body color is composed of purple and beige, the end of the tail is like a hand
Poisonous attributes, it looks like a huge purple cobra, with black stripes on its body, small head, and triangular eyes
Properties
All 898 images are from The Complete Pokemon Images Data Set in Kaggle with size 475x475. Each image is accompanied with corresponding pokemon name and its detailed description from Pokemon Wiki, English and Chinese captions are provided. Human efforts are also involved to revise.
How to use
from datasets import load_dataset
dataset = load_dataset("wanghaofan/pokemon-wiki-captions")
The dataset is formatted as below. For each row the dataset contains image, name_en, name_zh, text_en and text_zh keys. image is a varying size PIL jpeg, name is the name of pokemon, and text is the accompanying text caption. Only a train split is provided.
DatasetDict({
train: Dataset({
features: ['image', 'name_en', 'name_zh', 'text_en', 'text_zh'],
num_rows: 898
})
})
Citation
If you use this dataset in your work, please cite it as:
@misc{wanghaofan2022pokemon,
author = {Haofan Wang},
title = {Pokemon wiki captions},
year={2022},
howpublished= {\url{https://huggingface.co/datasets/wanghaofan/pokemon-wiki-captions/}}
}