近日，旷视科技研究团队推出了一款支持文档级OCR的多模态大模型Vary，该模型能一键将文档图片直接转换为Markdown格式。这一技术的出现，大大简化了以往繁琐的文档处理流程，包括文本识别、布局检测和排序、公式表格处理、文本清洗等多个步骤。
旷视科技的研究团队通过深度学习技术，实现了对文档图片的高精度识别，并能准确地将识别结果转化为Markdown格式。这款多模态大模型Vary不仅支持中英文，还能应对各种复杂的文档场景，如含有公式、表格等。
以往，想把一份文档图片转换为Markdown格式，需要经过多个步骤，而现在，只需输入一句话命令，Vary就能直接端到端输出文档结果。这一创新大大提高了工作效率，节省了人力和时间成本。
这项技术的出现，将进一步推动文档处理领域的技术革新，也为广大用户提供了一种全新的文档处理方式。未来，旷视科技将继续深化在人工智能领域的研发，为广大用户带来更多便捷、高效的技术产品。
英文翻译：
News Title: MegVII Launches Vary, a Multimodal Large Model That Converts Documents to Markdown with One Click
Keywords: MegVII, multimodal large model, document-level OCR, Markdown format
News Content:
Recently, the research team of MegVII has launched Vary, a multimodal large model that supports document-level OCR and can convert document images into Markdown format with one click. This technology greatly simplifies the complicated processing pipeline of document handling, including text recognition, layout detection and sorting, formula and table processing, text cleaning, and more.
Through deep learning technology, the MegVII research team has achieved high-precision recognition of document images and accurately transforms the recognition results into Markdown format. Vary supports both Chinese and English and can cope with various complex document scenarios, including those containing formulas, tables, etc.
In the past, converting a document image into Markdown format required multiple steps, such as text recognition, layout detection and sorting, formula and table processing, and text cleaning. Now, with Vary, users only need to input a command to directly output the document results, greatly improving work efficiency and saving manpower and time costs.
The emergence of this technology will further promote technological innovations in the field of document processing and provide users with a new way of document handling. In the future, MegVII will continue to deepen research and development in the field of artificial intelligence, bringing more convenient and efficient technology products to users.
【来源】https://www.qbitai.com/2023/12/109275.html
Views: 1