ucsahin/pubtables-detection-1500-samples

---
language:
- en
license: cdla-permissive-2.0
size_categories:
- 1K<n<10K
task_categories:
- object-detection
pretty_name: PubTables Dataset for Table Detection
dataset_info:
features:
- name: image
dtype: image
- name: objects
struct:
- name: bbox
sequence:
sequence: float64
- name: categories
dtype: string
splits:
- name: train
num_bytes: 228796393.0
num_examples: 1500
download_size: 224693659
dataset_size: 228796393.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
tags:
- Documents
- Tables
---
## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->
This dataset is a sampled version of [bsmock/pubtables-1m](https://huggingface.co/datasets/bsmock/pubtables-1m) consisting of 1500 images along with the PASCAL VOC formatted bounding boxes of multiple tables. It can be used for demonstration purposes in finetuning DETR and multimodal large language models.

### Dataset Sources [optional]

<!-- Provide the basic links for the dataset. -->

- **Repository:** [bsmock/pubtables-1m] (https://huggingface.co/datasets/bsmock/pubtables-1m)

## Uses

<!-- Address questions around how the dataset is intended to be used. -->
It can be used for finetuning DETR and TATR type models. Please check the original license in [microsoft/table-transformer](https://github.com/microsoft/table-transformer).

原始信息汇总

数据集概述

基本信息

• 名称: PubTables Dataset for Table Detection
• 语言: 英语
• 许可证: CDLA-Permissive-2.0
• 大小分类: 1K<n<10K
• 任务分类: 对象检测

数据集特征

• 特征:
  • image: 图像类型
  • objects: 结构化数据
    • bbox: 边界框，数据类型为浮点数序列
    • categories: 类别，数据类型为字符串

数据集分割

• 训练集:
  • 示例数量: 1500
  • 数据大小: 228796393.0 字节
  • 下载大小: 224693659 字节

配置

• 默认配置:
  • 数据文件:
    • 分割: 训练
    • 路径: data/train-*

数据集用途

• 用于微调DETR和TATR类型模型。