# A Two-Stage Method for Text Line Detection in Historical Documents

arXiv: https://arxiv.org/abs/1802.03345 · HTML: https://mlsanity.com/paper/1802.03345/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 9 Feb 2018.

## Built on by (most cited first, 26 of 26)

- [Document AI: Benchmarks, Models and Applications](https://mlsanity.com/paper/2111.08609/) (2021)
- [DAN: a Segmentation-free Document Attention Network for Handwritten Document Recognition](https://mlsanity.com/paper/2203.12273/) (2022)
- [Page Layout Analysis System for Unconstrained Historic Documents](https://mlsanity.com/paper/2102.11838/) (2021)
- [SPAN: a Simple Predict & Align Network for Handwritten Paragraph Recognition](https://mlsanity.com/paper/2102.08742/) (2021)
- [Unified Line and Paragraph Detection by Graph Convolutional Networks](https://mlsanity.com/paper/2203.09638/) (2022)
- [Digital Peter: Dataset, Competition and Handwriting Recognition Methods](https://mlsanity.com/paper/2103.09354/) (2021)
- [Robust Text Line Detection in Historical Documents: Learning and Evaluation Methods](https://mlsanity.com/paper/2203.12346/) (2022)
- [Large Scale Genealogical Information Extraction From Handwritten Quebec Parish Records](https://mlsanity.com/paper/2304.14044/) (2023)
- [The Socface Project: Large-Scale Collection, Processing, and Analysis of a Century of French Censuses](https://mlsanity.com/paper/2404.18706/) (2024)
- [Improving OCR Quality in 19th Century Historical Documents Using a Combined Machine Learning Based Approach](https://mlsanity.com/paper/2401.07787/) (2024)
- [You Actually Look Twice At it (YALTAi): using an object detection approach instead of region segmentation within the Kraken engine](https://mlsanity.com/paper/2207.11230/) (2022)
- [Nuremberg Letterbooks: A Multi-Transcriptional Dataset of Early 15th Century Manuscripts for Document Analysis](https://mlsanity.com/paper/2411.07138/) (2024)
- [From Fragment to One Piece: A Survey on AI-Driven Graphic Design](https://mlsanity.com/paper/2503.18641/) (2025)
- [Text line extraction using fully convolutional network and energy minimization](https://mlsanity.com/paper/2101.07370/) (2021)
- [Evaluation of a Region Proposal Architecture for Multi-task Document Layout Analysis](https://mlsanity.com/paper/2106.11797/) (2021)
- [Including Keyword Position in Image-based Models for Act Segmentation of Historical Registers](https://mlsanity.com/paper/2109.08477/) (2021)
- [Entry Separation using a Mixed Visual and Textual Language Model: Application to 19th century French Trade Directories](https://mlsanity.com/paper/2302.08948/) (2023)
- [Unsupervised learning of text line segmentation by differentiating coarse patterns](https://mlsanity.com/paper/2105.09405/) (2021)
- [OPAD: An Optimized Policy-based Active Learning Framework for Document Content Analysis](https://mlsanity.com/paper/2110.02069/) (2021)
- [Self-paced learning to improve text row detection in historical documents with missing labels](https://mlsanity.com/paper/2201.12216/) (2022)
- [Convolutional Neural Network for Early Pulmonary Embolism Detection via Computed Tomography Pulmonary Angiography](https://mlsanity.com/paper/2204.03204/) (2022)
- [Recognizing Handwriting Styles in a Historical Scanned Document Using Unsupervised Fuzzy Clustering](https://mlsanity.com/paper/2210.16780/) (2022)
- [EDocNet: Efficient Datasheet Layout Analysis Based on Focus and Global Knowledge Distillation](https://mlsanity.com/paper/2502.16541/) (2025)
- [FP-THD: Full page transcription of historical documents](https://mlsanity.com/paper/2601.17040/) (2026)
- [End-to-End Text Line Detection and Ordering](https://mlsanity.com/paper/2606.04166/) (2026)
- [Text region detection in historical astronomical diagrams](https://mlsanity.com/paper/2606.15886/) (2026)

## Abstract (arXiv)

This work presents a two-stage text line detection method for historical documents. Each detected text line is represented by its baseline. In a first stage, a deep neural network called ARU-Net labels pixels to belong to one of the three classes: baseline, separator or other. The separator class marks beginning and end of each text line. The ARU-Net is trainable from scratch with manageably few manually annotated example images (less than 50). This is achieved by utilizing data augmentation strategies. The network predictions are used as input for the second stage which performs a bottom-up clustering to build baselines. The developed method is capable of handling complex layouts as well as curved and arbitrarily oriented text lines. It substantially outperforms current state-of-the-art approaches. For example, for the complex track of the cBAD: ICDAR2017 Competition on Baseline Detection the F-value is increased from 0.859 to 0.922. The framework to train and run the ARU-Net is open source.