Hugging Face's logo

Datasets:
ymfl
/
pix2tex

Tasks:
Image-to-Text
Modalities:
Image
Text
Formats:
parquet
Languages:
English
Size:
100K - 1M
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
pix2tex
File size: 4,929 Bytes
a371edd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: latex
    dtype: string
  splits:
  - name: train
    num_bytes: 1638732362.625
    num_examples: 377163
  - name: validation
    num_bytes: 545374827.875
    num_examples: 125721
  - name: test
    num_bytes: 544914564.75
    num_examples: 125722
  download_size: 2468946641
  dataset_size: 2729021755.25
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
license: apache-2.0
task_categories:
- image-to-text
language:
- en
---
Here’s a completed dataset card for your **Pix2Tex** dataset based on the provided template:

---

# Pix2Tex

## Dataset Details

### Dataset Description

The **Pix2Tex Dataset** is a high-quality, curated dataset for training and evaluating Vision-Language Models (VLMs) capable of extracting LaTeX code from images of mathematical formulas. This dataset combines both **printed** and **handwritten** formula images, offering diverse and challenging samples for tasks involving LaTeX recognition.  

The dataset contains images paired with their LaTeX annotations, enabling researchers and developers to explore various use cases, including mathematical OCR (Optical Character Recognition), multimodal learning, and LaTeX generation from formula images.  

- **Curated by:** Anindya  
- **Language(s):** English (en)  
- **License:** Apache 2.0  

---

## Uses

### Direct Use

The **Pix2Tex Dataset** can be used for:  
- Training and evaluation of Vision-Language Models (VLMs).  
- Image-to-Text conversion for LaTeX generation.  
- Research in mathematical OCR and handwriting recognition.  
- Applications involving mathematical document digitization.  

### Out-of-Scope Use

The dataset should not be used for purposes unrelated to image-to-text conversion, LaTeX generation, or Vision-Language modeling tasks. Misuse of the dataset for malicious purposes, such as creating deceptive content, is strictly prohibited.

---

## Dataset Structure

### Features

The dataset contains the following features:  
- **`image`**: The image of the formula (either printed or handwritten).  
- **`latex`**: The corresponding LaTeX code representing the formula.  

### Splits

The dataset is split into three subsets for easier use in machine learning workflows:  

| Split       | Number of Examples | Size (bytes)       |
|-------------|---------------------|--------------------|
| **Train**   | 377,163             | 1,638,732,362.625 |
| **Validation** | 125,721          | 545,374,827.875   |
| **Test**    | 125,722             | 544,914,564.75    |

### Configurations

The dataset includes a default configuration with the following data files:
- Train: `data/train-*`
- Validation: `data/validation-*`
- Test: `data/test-*`

---

## Dataset Creation

### Curation Rationale

The **Pix2Tex Dataset** was created to address the need for diverse, high-quality datasets for training Vision-Language Models (VLMs) that can extract LaTeX from mathematical formula images. By including both printed and handwritten formulas, the dataset provides a robust benchmark for real-world applications.

### Source Data

#### Data Collection and Processing

The dataset was created by combining two existing datasets:  
- One dataset containing printed formula images with their LaTeX annotations.  
- Another dataset containing handwritten formula images with their LaTeX annotations.  

Both datasets were preprocessed to ensure consistency in column naming and annotation quality. The resulting dataset was split into training, validation, and test sets.

#### Who are the source data producers?

The source data was collected from publicly available datasets and curated for the specific purpose of LaTeX extraction tasks.

---

## Bias, Risks, and Limitations

### Risks and Limitations

- The dataset includes a mix of printed and handwritten formulas, which might introduce variability in performance for models trained solely on one type of input.  
- The handwritten formulas may contain noise due to variations in handwriting styles, potentially challenging OCR tasks.  

### Recommendations

Users are advised to preprocess the dataset and evaluate models on both printed and handwritten subsets to better understand their performance across different input types.

---

## Citation

If you use this dataset in your research or projects, please cite it as follows:  


## Glossary

- **Vision-Language Models (VLMs)**: Machine learning models that combine visual and textual inputs to perform tasks such as image-to-text generation.  
- **LaTeX**: A typesetting system commonly used for mathematical and scientific documents.  

---

## Dataset Card Contact

For any questions or feedback, feel free to contact:  
- **Hugging Face Profile:** [anindya-hf-2002](https://huggingface.co/anindya-hf-2002)