Writer Adaptation in Handwritten Text Recognition (HTR)

Contact: Marco Peer

Overview

The digitization of historical documents requires accurate transcription of handwritten text. Variations in writing styles and deteriorated conditions make recognition challenging.
This thesis explores writer identification and writer-specific style extraction to improve Handwritten Text Recognition (HTR). The goal is to enhance recognition accuracy by leveraging writer-specific characteristics.

To evaluate the proposed writer adaptation techniques, benchmark datasets such as IAM or CVL, widely used in Handwritten Text Recognition (HTR) research, should be employed. Additionally, the evaluation will incorporate a historical dataset known for its challenges, such as the Bullinger Dataset [1]. The existing methodology for writer adaptation in HTR systems primarily relies on meta-learning techniques [2] or involves incorporating networks to extract writer styles [3].

Objectives

• Review state-of-the-art writer adaptation methods in HTR
• Implement and evaluate methods on IAM, CVL, and Bullinger datasets
• Develop a writer adaptation algorithm for HTR
• Compare results using CER/WER metrics

Methodology

1. Literature Review – Study meta-learning, style extractor networks, and diffusion-based approaches
2. Implementation – Apply existing methods and develop a writer adaptation algorithm
3. Evaluation – Benchmark on IAM, CVL, and Bullinger datasets
4. Reporting – Final thesis, presentation, and optional publication

Skills Required

• Python programming
• Deep learning (preferably PyTorch)
• Interest in document analysis and historical handwriting

References

[1] A. Scius-Bertrand et al., Bullinger Dataset for Writer Adaptation (BullingerDB), Link
[2] A. Bhunia et al., MetaHTR: Towards writer-adaptive handwritten text recognition, CVPR, 2021
[3] Z. Wang and J. Du, Fast writer adaptation with style extractor network for handwritten text recognition, Neural Networks, 2022

Created: 01.09.2025