CIawevy/TextPecker-1.5M

---
license: apache-2.0
task_categories:
- image-to-text
dataset_info:
features:
- name: id
dtype: string
- name: images
list: image
- name: conversations
list:
- name: content
dtype: string
- name: role
dtype: string
- name: data_source
dtype: string
- name: class
dtype: string
- name: ori_bbox
list: string
splits:
- name: test
num_bytes: 986226411
num_examples: 1061
- name: train
num_bytes: 984872941236
num_examples: 1482028
download_size: 985226675892
dataset_size: 985859167647
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
- split: train
path: data/train-*
---

# TextPecker-1.5M: A Dataset for Training and evaluating TextPecker

This repository contains the **TextPecker-1.5M** dataset, a new benchmark proposed in the paper "[TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering](https://arxiv.org/abs/2602.20903)".

## Code and Project Page

The official implementation and project details for the TextPecker and TextPecker-1.5M dataset can be found on the GitHub repository:
[https://github.com/CIawevy/TextPecker](https://github.com/CIawevy/TextPecker)

## Sample Usage

You can easily load the TextPecker-1.5M dataset using the Hugging Face `datasets` library. The dataset is provided in two configurations: `train` and `test`

```python
from datasets import load_dataset

# Load the full TextPecker-1.5M dataset (includes train and test splits)
dataset = load_dataset("CIawevy/TextPecker-1.5M", "default")
train_data = dataset["train"]
test_data = dataset["test"]

# Load specific split directly (more efficient for practical usage)
train_data = load_dataset("CIawevy/TextPecker-1.5M", "default", split="train")
test_data = load_dataset("CIawevy/TextPecker-1.5M", "default", split="test")
```

For detailed instructions on installation, model download, evaluation, and running demos, please refer to the [GitHub repository](https://github.com/CIawevy/TextPecker).

## Citation

If you find this dataset useful for your research, please cite the accompanying paper:

```bibtex
@article{zhu2026TextPecker,
title = {TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering},
author = {Zhu, Hanshen and Liu, Yuliang and Wu, Xuecheng and Wang, An-Lan and Feng, Hao and Yang, Dingkang and Feng, Chao and Huang, Can and Tang, Jingqun and Bai, Xiang},
journal = {arXiv preprint arXiv:2602.20903},
year = {2026}
}
```