sungjun12/vlm_tsr_test_1

这是一个用于视觉语言模型(VLM)表格结构识别(TSR)测试的数据集的第一部分(共四部分)，主要包含韩语内容。数据集包含scene组50000~50004的样本，每个样本由三个文件组成：表格图片、包含正确HTML和元数据的JSON文件，以及用于渲染的HTML文件。该数据集用于评估和基准测试表格结构识别性能，评估指标包括基于树编辑距离的结构相似度(TEDS)、仅评估结构的TEDS-Structure，以及合并单元格预测准确度的Span Cell F1。

This is the first part (out of four) of a test dataset for Visual Language Model (VLM) Table Structure Recognition (TSR), primarily containing Korean content. The dataset includes samples from scene groups 50000~50004, with each sample consisting of three files: a table image, a JSON file containing correct HTML and metadata, and an HTML file for rendering. The dataset is used for evaluating and benchmarking table structure recognition performance, with evaluation metrics including Tree-Edit Distance-based structural similarity (TEDS), structure-only evaluation (TEDS-Structure), and Span Cell F1 for merged cell prediction accuracy.