|
|
--- |
|
|
license: cc-by-nc-sa-4.0 |
|
|
task_categories: |
|
|
- visual-question-answering |
|
|
- image-to-image |
|
|
language: |
|
|
- en |
|
|
- zh |
|
|
tags: |
|
|
- culture |
|
|
- hanfu |
|
|
size_categories: |
|
|
- 1K<n<10K |
|
|
--- |
|
|
# Dataset Card for Dataset Name |
|
|
|
|
|
<!-- Provide a quick summary of the dataset. --> |
|
|
Official Dataset for [Hanfu-Bench: A Multimodal Benchmark on Cross-Temporal Cultural Understanding and Transcreation](https://arxiv.org/abs/2506.01565) |
|
|
|
|
|
Official Code:[TemporalCulture](https://github.com/lizhou21/TemporalCulture) |
|
|
|
|
|
## Dataset License Statement |
|
|
This dataset is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license. |
|
|
|
|
|
### Core License Terms |
|
|
- **Attribution (BY):** Users must give appropriate credit to the original author(s) of the dataset. |
|
|
- **NonCommercial (NC):** This dataset is restricted to non-commercial use. Any use for commercial purposes is strictly prohibited. |
|
|
- **ShareAlike (SA):** Any derivative works based on this dataset must be distributed under the same CC BY-NC-SA 4.0 license. |
|
|
|
|
|
### Additional Restrictions |
|
|
To further clarify usage, this dataset is strictly **limited to academic research purposes**, defined as: |
|
|
- **Academic purposes**: Activities related to education, academic research, and scholarly communication, including but not limited to writing academic papers, evaluating models, and conducting experiments. |
|
|
- **Evaluation Only**: The data may only be used for evaluation purposes and not for training models or systems. |
|
|
- Non-academic non-commercial activities, such as personal projects or non-academic nonprofit initiatives, are not permitted. |
|
|
- All users must ensure compliance with relevant institutional and national laws and regulations during usage. |
|
|
|
|
|
|
|
|
## Dataset Description |
|
|
We introduce Hanfu-Bench, a novel, expert-curated multimodal dataset. |
|
|
In total, 496 sets of Hanfu are retained, comprising 1,192 images, with an average of 4.74 identified features per Hanfu. |
|
|
This dataset includes some figures |
|
|
(This dataset retains some images that were filtered out after annotation.) |
|
|
|
|
|
Hanfu-Bench comprises two core tasks: cultural visual understanding and cultural image transcreation. |
|
|
The former task examines temporal-cultural feature recognition based on single- or multi-image inputs through multiple-choice visual question answering, while the latter focuses on transforming traditional attire into modern designs through cultural element inheritance and modern context adaptation. |
|
|
Our evaluation shows that closed VLMs perform comparably to non-experts on visual cutural understanding but fall short by 10% to human experts, while open VLMs lags further behind non-experts. |
|
|
For the transcreation task, multi-faceted human evaluation indicates that the best-performing model achieves a success rate of only 42%. |
|
|
Our benchmark provides an essential testbed, revealing significant challenges in this new direction of temporal cultural understanding and creative adaptation. |
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
|
|
|
Overview of the dataset. |
|
|
|
|
|
1. meta-info.json |
|
|
|
|
|
``` |
|
|
Dataset({ |
|
|
"1000": { |
|
|
"source": "TaoBao", |
|
|
"img_list": [ |
|
|
"num1000_img1.jpg" |
|
|
], |
|
|
"gender": "female", |
|
|
"type": "传统汉服形制", |
|
|
"period": "唐朝", |
|
|
"xiu": "unsure", |
|
|
"jin": "对襟", |
|
|
"ling": "直领", |
|
|
"bottoms": "unsure", |
|
|
"outerwear": "披帛" |
|
|
}, |
|
|
}) |
|
|
``` |
|
|
|
|
|
2. merged-sivqa.json |
|
|
``` |
|
|
Dataset([ |
|
|
{ |
|
|
"question_id": "single_0", |
|
|
"question_type": "gender", |
|
|
"cloth_id": "1000", |
|
|
"img_list": [ |
|
|
"num1000_img1.jpg" |
|
|
], |
|
|
"base_question": "图片中的服饰通常适合什么性别？", |
|
|
"base_question_en": "Which gender is the clothing in this image typically suitable for, among the following options?", |
|
|
"choices": "A.男; B.女", |
|
|
"choices_en": "A.Male; B.Female", |
|
|
"answer": "B" |
|
|
}, |
|
|
]) |
|
|
``` |
|
|
|
|
|
3. merged-mivqa.json |
|
|
``` |
|
|
Dataset([ |
|
|
{ |
|
|
"question_meta": { |
|
|
"answer": { |
|
|
"image": "num1080_img4.jpg", |
|
|
"meta": { |
|
|
"gender": "female", |
|
|
"type": "汉元素服饰", |
|
|
"style": "unsure", |
|
|
"period": "unsure", |
|
|
"xiu": "unsure", |
|
|
"jin": "对襟", |
|
|
"ling": "圆领", |
|
|
"bottoms": "unsure", |
|
|
"outerwear": "" |
|
|
} |
|
|
}, |
|
|
"candidates": [ |
|
|
{ |
|
|
"image": "num166_img10.jpg", |
|
|
"meta": { |
|
|
"gender": "female", |
|
|
"type": "汉服改良版", |
|
|
"style": "TwoPiece", |
|
|
"period": "unsure", |
|
|
"xiu": "unsure", |
|
|
"jin": "对襟", |
|
|
"ling": "unsure", |
|
|
"bottoms": "马面裙", |
|
|
"outerwear": "" |
|
|
} |
|
|
}, |
|
|
{ |
|
|
"image": "num1169_img5.jpg", |
|
|
"meta": { |
|
|
"gender": "male", |
|
|
"type": "传统汉服形制", |
|
|
"style": "unsure", |
|
|
"period": "unsure", |
|
|
"xiu": "unsure", |
|
|
"jin": "大襟", |
|
|
"ling": "unsure", |
|
|
"bottoms": "裤", |
|
|
"outerwear": "" |
|
|
} |
|
|
}, |
|
|
{ |
|
|
"image": "num1310_img1.jpg", |
|
|
"meta": { |
|
|
"gender": "male", |
|
|
"type": "传统汉服形制", |
|
|
"style": "TwoPiece", |
|
|
"period": "unsure", |
|
|
"xiu": "unsure", |
|
|
"jin": "unsure", |
|
|
"ling": "unsure", |
|
|
"bottoms": "unsure", |
|
|
"outerwear": "甲胄" |
|
|
} |
|
|
} |
|
|
], |
|
|
"answer_img": "num1080_img4.jpg", |
|
|
"candidate_imgs": [ |
|
|
"num166_img10.jpg", |
|
|
"num1169_img5.jpg", |
|
|
"num1310_img1.jpg" |
|
|
], |
|
|
"question": "以下图片中的服饰属于汉元素服饰的是？", |
|
|
"question_type": "type", |
|
|
"question_formular": "type_t1" |
|
|
}, |
|
|
"question": "以下图片中的服饰属于汉元素服饰的是？", |
|
|
"options": [ |
|
|
"num1080_img4.jpg", |
|
|
"num166_img10.jpg", |
|
|
"num1169_img5.jpg", |
|
|
"num1310_img1.jpg" |
|
|
], |
|
|
"answer_idx": 0, |
|
|
"qid": "mivqa_0", |
|
|
"question_id": "multi_0", |
|
|
"question_type": "type" |
|
|
}, |
|
|
]) |
|
|
``` |
|
|
|
|
|
|
|
|
|
|
|
## Citation |
|
|
|
|
|
<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. --> |
|
|
|
|
|
``` |
|
|
@inproceedings{zhou-etal-2025-hanfu, |
|
|
title = "Hanfu-Bench: A Multimodal Benchmark on Cross-Temporal Cultural Understanding and Transcreation", |
|
|
author = "Zhou, Li and |
|
|
Yu, Lutong and |
|
|
Xie, Dongchu and |
|
|
Cheng, Shaohuan and |
|
|
Li, Wenyan and |
|
|
Li, Haizhou", |
|
|
editor = "Christodoulopoulos, Christos and |
|
|
Chakraborty, Tanmoy and |
|
|
Rose, Carolyn and |
|
|
Peng, Violet", |
|
|
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing", |
|
|
month = nov, |
|
|
year = "2025", |
|
|
address = "Suzhou, China", |
|
|
publisher = "Association for Computational Linguistics", |
|
|
url = "https://aclanthology.org/2025.emnlp-main.1251/", |
|
|
doi = "10.18653/v1/2025.emnlp-main.1251", |
|
|
pages = "24627--24649", |
|
|
ISBN = "979-8-89176-332-6", |
|
|
} |
|
|
``` |
|
|
|
|
|
|
|
|
|
|
|
## Dataset Card Contact |
|
|
If you have any questions regarding the dataset, please reach out to: |
|
|
lizhou21@cuhk.edu.cn |
