seongwon980/Nameonly_generated

---
license: mit
language:
- en
---
# Just Say the Name: Online Continual Learning with Category Names Only via Data Generation
We provide the dataset used for <a href="https://arxiv.org/abs/2403.10853" target="_new">Name-only continual learning</a>, generated using Stable Diffusion XL, DALL.E-2, CogView2, and DeepFloyd IF models.

## Disclaimer
This dataset is created solely for academic purposes. We minimized human intervention to ensure a fair comparison with the baseline methods discussed in our paper. Despite our efforts, the extensive size of the dataset prevented us from examining each image individually.

Therefore, we emphasize that this dataset should only be used for academic purposes, and we disclaim any issues arising from improper use. Specifically, this dataset is recommended solely for assessing the performance of classification models derived from datasets generated by generative models.

## Dataset Composition
We have constructed generative model versions of three publicly available datasets: PACS[1], DomainNet[2], and CCT[2]. All datasets will be released sequentially. Additionally, we plan to upload all datasets that have been processed using our proposed RMD-based ensemble method.

1. PACS dataset from generative model
2. DomainNet dataset from generative model
3. CCT dataset from generative model

## Download the dataset
clone this repository by following the commands below.
```sh
git clone https://huggingface.co/datasets/seongwon980/Nameonly_generated
```

After downloading the dataset, you may check the directories and the expected outcome is as below.
```sh
cd Nameonly_generated/raw_datasets/PACS_sdxl
ls
dog elephant giraffe guitar horse house person
```

## Related work that uses this dataset
<ul>
<li>
<a href="https://arxiv.org/abs/2403.10853">
Just Say the Name: Online Continual Learning with Category Names Only via Data Generation
</a>
<br>
<a href="https://dbd05088.github.io/" target="_new">
Minhyuk Seo
</a>,
<a href="https://digantamisra98.github.io/" target="_new">
Diganta Misra
</a>,
<a href="https://huggingface.co/datasets/seongwon980/Nameonly_generated" target="_new">
Seongwon Cho
</a>,
<a href="https://huggingface.co/datasets/seongwon980/Nameonly_generated" target="_new">
Minjae Lee
</a>,
<a href="http://ppolon.github.io/" target="_new">
Jonghyun Choi
</a>
<br>
arXiv 2024
</li>
</ul>

## References

[1] Li, Da, et al. "Deeper, broader and artier domain generalization." ICCV, 2017.

[2] Peng, Xingchao, et al. "Moment matching for multi-source domain adaptation." ICCV, 2019.

[3] Beery, Sara, Grant Van Horn, and Pietro Perona. "Recognition in terra incognita." ECCV, 2018.

## Citation
If you find this repository useful, please cite this repository.
```
@article{seo2024just,
title={Just Say the Name: Online Continual Learning with Category Names Only via Data Generation},
author={Seo, Minhyuk and Misra, Diganta and Cho, Seongwon and Lee, Minjae and Choi, Jonghyun},
journal={arXiv preprint arXiv:2403.10853},
year={2024}
}
```

原始信息汇总

数据集概述

数据集名称

• Just Say the Name: Online Continual Learning with Category Names Only via Data Generation

数据集用途

• 用于学术研究，特别是评估由生成模型生成的数据集上分类模型的性能。

数据集生成方法

• 使用Stable Diffusion XL, DALL.E-2, CogView2, 和 DeepFloyd IF模型生成。

数据集构成

• 包含三个由生成模型处理过的公开数据集版本：
  1. PACS dataset from generative model
  2. DomainNet dataset from generative model
  3. CCT dataset from generative model

数据集下载

• 通过以下命令克隆仓库获取数据集： sh git clone https://huggingface.co/datasets/seongwon980/Nameonly_generated

数据集目录结构示例

• 示例目录结构如下： sh cd Nameonly_generated/raw_datasets/PACS_sdxl ls dog elephant giraffe guitar horse house person

引用信息

• 引用格式：

  @article{seo2024just, title={Just Say the Name: Online Continual Learning with Category Names Only via Data Generation}, author={Seo, Minhyuk and Misra, Diganta and Cho, Seongwon and Lee, Minjae and Choi, Jonghyun}, journal={arXiv preprint arXiv:2403.10853}, year={2024} }