metadata
dataset_info:
features:
- name: id
dtype: string
- name: created_at
dtype: string
- name: prompt
dtype: string
- name: negative_prompt
dtype: string
- name: likes
dtype: int64
- name: sampler
dtype: string
- name: height
dtype: int64
- name: steps
dtype: int64
- name: width
dtype: int64
- name: cursor
dtype: int64
- name: url
dtype: string
- name: cfg_scale
dtype: float64
- name: model
dtype: string
splits:
- name: train
num_bytes: 248764537
num_examples: 256224
download_size: 54319285
dataset_size: 248764537
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
tags:
- image generation
- negative prompts
- stable-diffusion
pretty_name: NegOpt Full
language:
- en
size_categories:
- 100K<n<1M
task_categories:
- text-to-image
- text-generation
This is the dataset constructed in and used to fine-tune the models proposed in our paper Optimizing Negative Prompts for Enhanced Aesthetics and Fidelity in Text-To-Image Generation.
If you find this dataset useful, please cite us here:
@article{ogezi2024optimizing,
title={Optimizing Negative Prompts for Enhanced Aesthetics and Fidelity in Text-To-Image Generation},
author={Ogezi, Michael and Shi, Ning},
journal={arXiv preprint arXiv:2403.07605},
year={2024}
}