amishor/DeepSeek-OCR
收藏资源简介：
DeepSeek-OCR数据集是一个包含2200个高质量图文对的多模态数据集，来源于《DeepSeek-OCR: Contexts Optical Compression for Ultra-Long Document Understanding》技术报告。这些图文对覆盖了学术论文、财务报告、教科书、幻灯片、报纸、图表、化学公式和几何图形等多种文档类型，旨在支持端到端的OCR、版面解析以及视觉语言预训练研究。数据集主要包含中文和英文两种语言，并支持大约100种语言的子集。
The DeepSeek-OCR Dataset is a high-quality multimodal dataset containing 2,200 image-text pairs derived from the technical report DeepSeek-OCR: Contexts Optical Compression for Ultra-Long Document Understanding. These pairs cover various document types including academic papers, financial reports, textbooks, slides, newspapers, charts, chemical formulas, and geometric figures, supporting end-to-end OCR, layout parsing, and vision-language pretraining research. The dataset primarily consists of Chinese and English languages and supports a subset of about 100 languages.
