Doclayout-yolo improves document layout detection using synthetic data and page axis attention
DocLayout-YOLO: Axis-Attentive Document Layout Detector with Mesh-Guided Synthetic Pretraining
Computer Vision and Pattern Recognition
Summary
Understanding the layout of documents, like finding titles, paragraphs, or images, is important for many digital tools, but it's hard when documents are very different. The authors created DocLayout-YOLO, a fast system that learns from a large set of computer-generated documents to better detect layout parts in many kinds of documents. They also designed a clever way for the system to pay attention to horizontal and vertical patterns on the page, helping it recognize layouts more accurately. Tests showed their method works well on various document types and runs quickly enough for real-time use.
What this means in practice
- •For document processing teams: Integrate DocLayout-YOLO to enhance accuracy and speed in detecting document regions across diverse document types.
- •For digital archiving professionals: Use the system to automate layout recognition for large scale scanning and organizing of historical or varied documents.
Authors
Zhiyuan Zhao, Hengrui Kang, Guang Liang, Bin Wang, Conghui He
Abstract
Accurate and efficient document layout analysis is critical for modern document parsing systems. However, although the YOLO framework represents a leading paradigm for the real-time detector, its adaptation to heterogeneous documents remains underexplored. To address this gap, we propose DocLayout-YOLO, a document-specialized real-time detector that integrates diverse synthetic document pretraining and page-axis context modeling. From the data perspective, we introduce MeshLayout-1M, a synthetic pretraining corpus generated by the proposed Mesh-candidate BestFit algorithm, which provides large-scale and diverse document layouts for pretraining and improves adaptation to heterogeneous document scenarios. From the model perspective, we propose Page-Axis Layout Attention (PALA), which captures long-range dependencies along the horizontal and vertical page axes while remaining computationally efficient. Extensive experiments demonstrate that DocLayout-YOLO achieves state-of-the-art performance on four heterogeneous document benchmarks, obtaining 72.4%/75.5%/96.0%/73.1% mAP50:95 on D4LA, M6Doc, HJDataset, and PRIMA-LAD, respectively. Compared with existing real-time detectors, DocLayout-YOLO establishes a superior accuracy-efficiency trade-off in document layout analysis. Code, model, and data are available at https://github.com/opendatalab/DocLayout-YOLO/tree/release-v2.