Datasets:
U-RxnDiagram-15k 数据集
数据集概述
U-RxnDiagram-15k 数据集是一个专门用于化学反应图表解析的大规模数据集,包含从科学文献PDF中提取的化学反应图像及其详细标注。该数据集旨在支持化学信息学、文档分析和计算机视觉领域的研究。
数据集统计信息
- 总图像数量: 15,400 张
- 训练集: 15,000 张图像
- 测试集: 400 张图像
- 总反应数量: 48,255 个反应
- 训练集: 45,426 个反应
- 测试集: 2,829 个反应
- 数据来源: 科学文献PDF文件
- 图像格式: PNG
- 标注总数: 约 165,468 个标注实例
数据集结构
U-RxnDiagram-15k/
├── train_set/
│ ├── ground_truth.json # 训练集标注文件
│ └── images/ # 训练集图像目录
└── test_set/
├── ground_truth.json # 测试集标注文件
└── images/ # 测试集图像目录
标注类别定义
数据集定义了4个主要类别,每个类别包含多个细分的属性:
1. Structure (分子结构) - category_id: 1
- P-reactants: 反应物分子
- P-product: 产物分子
- P-reaction conditions: 反应条件
2. Text (文本) - category_id: 2
- T-reaction conditions: 反应条件文本
- T-reactants: 反应物文本
- T-product: 产物文本
3. Identifier (标识符) - category_id: 3
- 化学标识符和编号
4. Supplement (补充信息) - category_id: 4
- 其他补充信息
标注统计
训练集
| 属性类型 | 标注数量 | 占比 |
|---|---|---|
| T-reaction condition | 56,377 | 35.84% |
| P-reactant | 31,779 | 20.20% |
| P-product | 30,808 | 28.79% |
| T-reactant | 6,433 | 6.02% |
| T-product | 3,804 | 2.42% |
| P-reaction condition | 6,230 | 3.96% |
测试集
| 属性类型 | 标注数量 | 占比 |
|---|---|---|
| T-reaction condition | 3,011 | 36.92% |
| P-reactant | 1,521 | 18.65% |
| P-product | 2,348 | 28.79% |
| T-reactant | 491 | 6.02% |
| T-product | 388 | 4.76% |
| P-reaction condition | 397 | 4.87% |
数据格式
图像文件命名
图像文件名已进行哈希处理(SHA-256 前 8 位)。
例如:a1b2c3d4.png。
标注文件格式 (ground_truth.json)
标注文件采用COCO格式,包含以下主要字段:
{
"licenses": [...],
"info": {
"description": "A dataset for chemical visual diagram analysis",
"version": "v1",
"year": "2025"
},
"categories": [
{"id": 1, "name": "structure"},
{"id": 2, "name": "text"},
{"id": 3, "name": "identifier"},
{"id": 4, "name": "supplement"}
],
"images": [
{
"id": 2,
"class": "figure",
"confidence": 0.9148465991020203,
"bbox": [x1, y1, x2, y2],
"original_id": 0,
"width": 1008.7104797363281,
"height": 471.88232421875,
"file_name": "ays765k9.png",
"bboxes": [
{
"id": 0,
"bbox": [x, y, width, height],
"category_id": 1,
"category": "P",
"attribute": "P-reactants",
"region_id": ["akzkPsql"]
}
]
}
]
}
标注字段说明
- id: 标注唯一标识符
- bbox: 边界框坐标 [x1, y1, x2, y2] 用于图像级bbox,[x, y, width, height] 用于bboxes列表
- category_id: 类别ID (1-4)
- category: 类别简称 (P=结构, T=文本)
- attribute: 具体属性名称
- region_id: 区域标识符列表
使用场景
该数据集适用于以下研究任务:
- 化学结构识别: 识别和定位化学反应中的分子结构
- 文本信息提取: 从化学图表中提取文本信息
- 反应条件分析: 识别和分析反应条件
- 文档理解: 理解科学文献中的化学信息
- 多模态学习: 结合视觉和文本信息进行化学分析
数据质量
- 所有图像均来自高质量的科学文献
- 标注经过专业验证,确保准确性
- 包含多种化学反应类型和复杂度的样本
- 支持细粒度的化学信息分析
引用
如果您在研究中使用了此数据集,请引用相关论文:
@article{song2025rxncaption,
title={RxnCaption: Reformulating Reaction Diagram Parsing as Visual Prompt Guided Captioning},
author={Song, Jiahe and Wang, Chuang and Jiang, Bowen and Wang, Yinfan and Zheng, Hao and Wei, Xingjian and Liu, Chengjin and Nie, Rui and Gao, Junyuan and Sun, Jiaxing and others},
journal={arXiv preprint arXiv:2511.02384},
year={2025}
}
许可证
本数据集采用 Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0) 许可证。
使用条款
您可以自由地:
- 共享 — 以任何媒介或格式复制和重新分发材料
- 改编 — 对材料进行混合、转换和构建
在以下条件下:
- 署名 — 您必须给出适当的署名,提供指向本许可协议的链接,同时标明是否对原始作品作了修改。您可以用任何合理的方式来署名,但是不得以任何方式暗示许可方认可您或您的使用。
- 非商业性使用 — 您不得将本材料用于商业目的。未经许可方明确许可,禁止商业使用。
引用格式
如果您在研究中使用了本数据集,请按以下格式引用:
## 联系方式
如有问题或建议,请联系 [email protected]
---
*最后更新: 2025年*