wanderkid commited on
Commit
fd6f02a
1 Parent(s): 3b6b0cf

Upload UniMER-1M and UniMER-Test Dataset

Browse files
Files changed (3) hide show
  1. README.md +118 -3
  2. UniMER-1M.zip +3 -0
  3. UniMER-test.zip +3 -0
README.md CHANGED
@@ -1,3 +1,118 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # UniMER 数据集
2
+ ## 简介
3
+ UniMER数据集是专门为通用数学表达式识别(MER)发布的数据集。它包含了真实全面的UniMER-1M训练集,拥有超过一百万个代表广泛和复杂数学表达式的实例,以及精心设计的UniMER测试集,用于在真实世界场景中评估MER模型。数据集详情如下:
4
+
5
+ - **UniMER-1M 训练集:**
6
+ - 总样本数:1,061,791
7
+ - 组成:简洁与复杂、扩展公式表达式的平衡融合
8
+ - 目标:帮助训练鲁棒性强、高精度的MER模型,增强识别准确性和模型泛化能力
9
+
10
+ - **UniMER 测试集:**
11
+ - 总样本数:23,757,分为四种表达式类型:
12
+ - 简单印刷表达式(SPE):6,762 个样本
13
+ - 复杂印刷表达式(CPE):5,921 个样本
14
+ - 屏幕截图表达式(SCE):4,774 个样本
15
+ - 手写表达式(HWE):6,332 个样本
16
+ - 目的:为MER模型提供一个全面的评估平台,以准确评估真实场景下各类公式识别能力
17
+
18
+ ## 视觉数据样本
19
+ ![UniMER-测试集](https://github.com/opendatalab/UniMERNet/assets/69186975/7301df68-e14c-4607-81bc-b6ee3ba1780b)
20
+
21
+ ## 数据统计
22
+ | **数据集** | **子集** | **来源** | **样本数量** |
23
+ |:-----------:|:-------:|:-------------------------------------------:|:------------:|
24
+ | UniMER-1M | | Pix2tex 训练集 | 158,303 |
25
+ | | | Arxiv † | 820,152 |
26
+ | | | CROHME 训练集 | 8,834 |
27
+ | | | HME100K 训练集 ‡ | 74,502 |
28
+ | UniMER-测试集 | SPE | Pix2tex 验证集 | 6,762 |
29
+ | | CPE | Arxiv † | 5,921 |
30
+ | | SCE | PDF 截图 † | 4,742 |
31
+ | | HWE | CROHME & HME100K | 6,332 |
32
+
33
+ † 表示由我们团队收集、处理和注释的数据。
34
+ ‡ 由于版权合规,请手动下载此部分数据集:[HME100K 数据集](https://ai.100tal.com/dataset)。
35
+
36
+ ## 致谢
37
+ 我们对[Pix2tex](https://github.com/lukas-blecher/LaTeX-OCR), [CROHME](https://www.cs.rit.edu/~rlaz/files/CROHME+TFD%E2%80%932019.pdf)和[HME100K](https://github.com/tal-tech/SAN) 数据集的创建者表示感谢。他们的基础工作对 UniMER 数据集的构建及发布做出了重大贡献。
38
+
39
+ ## 引用
40
+ ```text
41
+ @misc{wang2024unimernet,
42
+ title={UniMERNet: A Universal Network for Real-World Mathematical Expression Recognition},
43
+ author={Bin Wang and Zhuangcheng Gu and Chao Xu and Bo Zhang and Botian Shi and Conghui He},
44
+ year={2024},
45
+ eprint={2404.15254},
46
+ archivePrefix={arXiv},
47
+ primaryClass={cs.CV}
48
+ }
49
+
50
+ @misc{conghui2022opendatalab,
51
+ author={He, Conghui and Li, Wei and Jin, Zhenjiang and Wang, Bin and Xu, Chao and Lin, Dahua},
52
+ title={OpenDataLab: Empowering General Artificial Intelligence with Open Datasets},
53
+ howpublished = {\url{https://opendatalab.com}},
54
+ year={2022}
55
+ }
56
+ ```
57
+
58
+ ---
59
+
60
+ # UniMER Dataset
61
+ ## Introduction
62
+ The UniMER dataset is a specialized collection curated to advance the field of Mathematical Expression Recognition (MER). It encompasses the comprehensive UniMER-1M training set, featuring over one million instances that represent a diverse and intricate range of mathematical expressions, coupled with the UniMER Test Set, meticulously designed to benchmark MER models against real-world scenarios. The dataset details are as follows:
63
+
64
+ - **UniMER-1M Training Set:**
65
+ - Total Samples: 1,061,791 Latex-Image pairs
66
+ - Composition: A balanced mix of concise and complex, extended formula expressions
67
+ - Aim: To train robust, high-accuracy MER models, enhancing recognition precision and generalization
68
+
69
+ - **UniMER Test Set:**
70
+ - Total Samples: 23,757, categorized into four types of expressions:
71
+ - Simple Printed Expressions (SPE): 6,762 samples
72
+ - Complex Printed Expressions (CPE): 5,921 samples
73
+ - Screen Capture Expressions (SCE): 4,774 samples
74
+ - Handwritten Expressions (HWE): 6,332 samples
75
+ - Purpose: To provide a thorough evaluation of MER models across a spectrum of real-world conditions
76
+
77
+
78
+ ## Visual Data Samples
79
+ ![UniMER-Test](https://github.com/opendatalab/UniMERNet/assets/69186975/7301df68-e14c-4607-81bc-b6ee3ba1780b)
80
+
81
+ ## Data Statistics
82
+ | **Dataset** | **Sub** | **Source** | **Sample Size** |
83
+ |:-----------:|:-------:|:-------------------------------------------:|:---------------:|
84
+ | UniMER-1M | | Pix2tex Train | 158,303 |
85
+ | | | Arxiv † | 820,152 |
86
+ | | | CROHME Train | 8,834 |
87
+ | | | HME100K Train ‡ | 74,502 |
88
+ | UniMER-Test | SPE | Pix2tex Validation | 6,762 |
89
+ | | CPE | Arxiv † | 5,921 |
90
+ | | SCE | PDF Screenshot † | 4,742 |
91
+ | | HWE | CROHME & HME100K | 6,332 |
92
+
93
+ † Indicates data collected, processed, and annotated by our team.
94
+ ‡ For copyright compliance, please manually download this dataset portion: [HME100K dataset](https://ai.100tal.com/dataset).
95
+
96
+ ## Acknowledgements
97
+ We would like to express our gratitude to the creators of the [Pix2tex](https://github.com/lukas-blecher/LaTeX-OCR), [CROHME](https://www.cs.rit.edu/~rlaz/files/CROHME+TFD%E2%80%932019.pdf), and [HME100K](https://github.com/tal-tech/SAN) datasets. Their foundational work has significantly contributed to the development of the UniMER dataset.
98
+
99
+
100
+ ## Citations
101
+
102
+ ```text
103
+ @misc{wang2024unimernet,
104
+ title={UniMERNet: A Universal Network for Real-World Mathematical Expression Recognition},
105
+ author={Bin Wang and Zhuangcheng Gu and Chao Xu and Bo Zhang and Botian Shi and Conghui He},
106
+ year={2024},
107
+ eprint={2404.15254},
108
+ archivePrefix={arXiv},
109
+ primaryClass={cs.CV}
110
+ }
111
+
112
+ @misc{conghui2022opendatalab,
113
+ author={He, Conghui and Li, Wei and Jin, Zhenjiang and Wang, Bin and Xu, Chao and Lin, Dahua},
114
+ title={OpenDataLab: Empowering General Artificial Intelligence with Open Datasets},
115
+ howpublished = {\url{https://opendatalab.com}},
116
+ year={2022}
117
+ }
118
+ ```
UniMER-1M.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18119240c1d9e186fd962ab9dc7e0e57314ef77fbed0a01152c2852dfb58f6bd
3
+ size 2022836719
UniMER-test.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69cf3eab39433ac67e0e0996bb0dcf1f26b24906264285a107cd1e5f3a65c7e8
3
+ size 147672072