Handwriting Recognition of Historical Documents with Few Labeled Data

Abstract

Historical documents present many challenges for offline handwriting recognition systems, among them, the segmentation and labeling steps. Carefully annotated textlines are needed to train an HTR system. In some scenarios, transcripts are only available at the paragraph level with no text-line information. In this work, we demonstrate how to train an HTR system with few labeled data. Specifically, we train a deep convolutional recurrent neural network (CRNN) system on only 10% of manually labeled text-line data from a dataset and propose an incremental training procedure that covers the rest of the data. Performance is further increased by augmenting the training set with specially crafted multiscale data. We also propose a model-based normalization scheme which considers the variability in the writing scale at the recognition phase. We apply this approach to the publicly available READ dataset 1 . Our system achieved the second best result during the ICDAR2017 competition [1].

A set of benchmarks for Handwritten Text Recognition on historical documents

“…See all the available details in [26]. The second row corresponds to a system that used 13 CNN layers and 3 layers of BLSTM, along with a word 2-gram for decoding [31]. The results of the here proposed baseline system are also shown in Table 9.…”

Boosting Offline Handwritten Text Recognition in Historical Documents With Few Labeled Lines

“…Similar to [32], in [36] they also apply some elastic distortions to the original images. In [4] the authors improve the performance by augmenting the training set with specially crafted multiscale data. They also propose a model-based normalization scheme that considers the variability in the writing scale at the recognition phase.…”

Improving Text Recognition using Optical and Language Model Writer Adaptation

“…Recently, as part of the ICFHR 2018 READ competition [8], most of the participants proposed an optical model composed of a Convolutional Neural Network (CNN) and Bidirectional Long Short Term Memory (BLSTM) layers as in [15] and [16]. One of them was using Multi-dimensional LSTM (MDLSTM) [17], which has provided good performance when trained on a generic large data set, but has shown difficulty in carrying the writer adaptation process with few samples.…”