Researchers have developed a pipeline that uses a vision large language model (LLM) combined with classical computer vision to extract structured diffraction data from images in scientific papers. This method can interpret various figure formats, locate diffraction plots, and accurately trace curves, transforming static images into reusable experimental data. Applied to 3,954 articles on covalent organic frameworks (COFs), the pipeline generated over 11,000 calibrated diffraction curves, which were released in the open database COF-PEAK. This approach allows for more detailed analysis, such as tracking changes in peak prominence, which is not typically possible with the limited peak angles usually reported in papers. AI
IMPACT Enables more comprehensive data extraction from scientific literature, potentially accelerating materials science research.
RANK_REASON The item describes a new method for extracting data from scientific papers, presented as a research paper on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →