SJY-1995/GeoTikz-Base

--- dataset_id: GeoTikz-Base license: cc-by-nc-4.0 language: - en tags: - image-to-tikz - geometry - mathematical-diagrams - tikz - plane-geometry - analytic-geometry - geo-tikz pretty_name: GeoTikz-Base size_category: large task_categories: - image-to-text --- # GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning [CVPR2026] ## Table of Contents 1. [Dataset Summary](#dataset-summary) 2. [Dataset Structure](#dataset-structure) 3. [Dataset Creation](#dataset-creation) 4. [Dataset Distribution](#dataset-distribution) 5. [Citation](#citation) --- ## Dataset Summary **GeoTikz-Base** is a 2.45M **image-to-TikZ dataset**. This dataset provides paired data consisting of original images and their corresponding TikZ code, serving as a foundational resource for research on geometric perception, mathematical reasoning, TikZ code synthesis, etc. ## Dataset Structure Each sample in GeoTikz-Base is composed of: 1. `image`: image bytes 2. `question`: query to generate tikz code 3. `response`: generated tikz code ## Dataset Creation The dataset is built using two core innovative strategies: 1. **Iterative Self-Refinement** 2. **Localized Transformation Strategy** Please refer to our paper for details. ## Dataset Distribution The distribution of GeoTikz-Base centers on **Plane Geometry** and **Analytic Geometry** as the primary components, also incorporating supplementary domains including Algebra, Calculus, and other mathematical fields, achieving full coverage of mainstream mathematical figures. <img src="dataset-distribution.jpg" width="60%" alt="GeoTikz-Base Dataset Domain Distribution"> ## Citation Please cite the paper if you use our data, model or code: ```bibtex @inproceedings{ geotikzbridge, title={GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning}, author={Jiayin Sun and Caixia Sun and Boyu Yang and Hailin Li and Xiao Chen and Yi Zhang and Errui Ding and Liang Li and Chao Deng and Junlan Feng}, booktitle={2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition}, year={2026} }数据集ID：GeoTikz-Base 许可证：CC-BY-NC-4.0 语言：英语 标签： - 图像转TikZ（image-to-tikz） - 几何学（geometry） - 数学图表（mathematical-diagrams） - TikZ（tikz） - 平面几何学（plane-geometry） - 解析几何学（analytic-geometry） - GeoTikz（geo-tikz） 友好名称：GeoTikz-Base 规模类别：大规模 任务类别：图像转文本（image-to-text） # GeoTikzBridge：推动几何感知与推理的多模态代码生成研究 [CVPR 2026] ## 目录 1. [数据集概述](#dataset-summary) 2. [数据集结构](#dataset-structure) 3. [数据集构建](#dataset-creation) 4. [数据集分布](#dataset-distribution) 5. [引用信息](#citation) --- ## 数据集概述 **GeoTikz-Base** 是一个包含245万条样本的**图像转TikZ**数据集。该数据集提供原始图像与对应TikZ代码的配对数据，可作为几何感知、数学推理、TikZ代码合成等领域研究的基础资源。 ## 数据集结构 GeoTikz-Base中的每个样本由以下三部分组成： 1. `image`：图像字节流 2. `question`：生成TikZ代码的查询指令 3. `response`：生成的TikZ代码 ## 数据集构建 本数据集通过两项核心创新策略构建： 1. **迭代自我精调（Iterative Self-Refinement）** 2. **局部变换策略（Localized Transformation Strategy）** 详细内容请参阅本团队的研究论文。 ## 数据集分布 GeoTikz-Base的数据集分布以**平面几何学**与**解析几何学**为核心主体，同时涵盖代数、微积分等其他数学领域作为补充，实现了主流数学图形的全覆盖。  ## 引用信息 若您使用本数据集、模型或代码，请引用以下论文： bibtex @inproceedings{ geotikzbridge, title={GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning}, author={Jiayin Sun and Caixia Sun and Boyu Yang and Hailin Li and Xiao Chen and Yi Zhang and Errui Ding and Liang Li and Chao Deng and Junlan Feng}, booktitle={2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition}, year={2026} }