Reproduce “DocLayout-YOLO: Axis-Attentive Document Layout Detector with Mesh-Guided Synthetic Pretraining” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhiyuan Zhao, Hengrui Kang, Guang Liang, Bin Wang, Conghui He
DocLayout-YOLO adapts YOLOv12 for document layout analysis by pretraining on synthetic pages and replacing Area Attention with parallel horizontal and vertical context paths. Its MeshLayout-1M corpus contains 1 million pages, while PALA connects features across full rows and columns. On M6Doc, X-scale achieves 75.5% mAP50:95 and 6.9 ms latency versus 74.0% and 8.5 ms for YOLOv12-X, under specified test conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.