LayoutParser: A Unified Toolkit for Deep Learning Based Document Image Analysis

Recent advances in document image analysis (DIA) have been primarily driven by the application of neural networks. Ideally, research outcomes could be easily deployed production and extended for further investigation. However, various factors like loosely organized codebases sophisticated model configurations complicate easy reuse important innovations a wide audience. Though there on-going efforts to improve reusability simplify deep learning (DL) development disciplines natural language processing computer vision, none them are optimized challenges domain DIA. This represents major gap existing toolkit, as DIA is central academic across range social sciences humanities. paper introduces LayoutParser, an open-source library streamlining usage DL applications. The core LayoutParser comes with set simple intuitive interfaces applying customizing models layout detection, character recognition, many other tasks. To promote extensibility, also incorporates community platform sharing both pre-trained full digitization pipelines. We demonstrate that helpful lightweight large-scale pipelines real-word use cases. publicly available at https://layout-parser.github.io.