tech4humans/br-doc-extraction

巴西文档结构提取数据集包含1218张巴西身份证件（CNH - 国家驾驶执照，RG - 一般注册）和发票（NF - 财政发票）的图像。每张图像都配有一个用户定义的JSON模式（作为“前缀”）和相应的结构化数据提取（作为JSON字符串格式的“后缀”）。该数据集的主要目标是帮助微调视觉语言模型，以从各种巴西文档图像中提取结构化信息。

The Brazilian Document Structure Extraction dataset contains 1218 images of Brazilian identification documents (CNH - National Drivers License, RG - General Registration) and invoices (NF - Nota Fiscal). Each image is paired with a user-defined JSON schema (as a prefix) and the corresponding structured data extraction (as a suffix in JSON string format). The primary goal of this dataset is to facilitate the fine-tuning of Vision-Language Models (VLMs) for extracting structured information from diverse Brazilian document images.