Recurrent neural networks are powerful sequence learners. They are able to incorporate context information in a flexible way, and are robust to localised distortions of the input data. These properties make them well suited to sequence labelling, where input sequences are transcribed with streams of labels. The aim of this thesis is to advance the state-of-the-art in supervised sequence labelling with recurrent networks. Its two main contributions are (1) a new type of output layer that allows recurrent networks to be trained directly for sequence labelling tasks where the alignment between the inputs and the labels is unknown, and (2) an extension of the long short-term memory network architecture to multidimensional data, such as images and video sequences.

https://mediatum.ub.tum.de/doc/673554/file.pdf

Supervised Sequence Labelling with Recurrent Neural Networks

Offline handwriting recognition—the automatic transcription of images of handwritten text—is a challenging task that combines computer vision with sequence learning. In most systems the two elements are handled separately, with sophisticated preprocessing techniques used to extract the image features and sequential models such as HMMs used to provide the transcriptions. By combining two recent innovations in neural networks—multidimensional recurrent neural networks and connectionist temporal classification—this paper introduces a globally trained offline handwriting recogniser that takes raw pixel data as input. Unlike competing systems, it does not require any alphabet specific preprocessing, and can therefore be used unchanged for any language. Evidence of its generality and power is provided by data from a recent international Arabic recognition competition, where it outperformed all entries (91.4% accuracy compared to 87.2% for the competition winner) despite the fact that neither author understands a word of Arabic.

/pdf/offline-handwriting-recognition-with-multidimensional-4710cm0ru0.pdf

Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks

Using recurrent neural network models for early detection of heart failure onset.

The automatic recognition of text on scanned images has enabled many applications such as searching for words in large volumes of documents, automatic sorting of postal mail, and convenient editing of previously printed documents. The domain of handwriting in the Arabic script presents unique technical challenges and has been addressed more recently than other domains. Many different methods have been proposed and applied to various types of images. This paper provides a comprehensive review of these methods. It is the first survey to focus on Arabic handwriting recognition and the first Arabic character recognition survey to provide recognition rates and descriptions of test data for the approaches discussed. It includes background on the field, discussion of the methods, and future research directions.

Offline Arabic handwriting recognition: a survey

Recurrent neural networks (RNNs) with Long Short-Term memory cells currently hold the best known results in unconstrained handwriting recognition. We show that their performance can be greatly improved using dropout - a recently proposed regularization method for deep architectures. While previous works showed that dropout gave superior performance in the context of convolutional networks, it had never been applied to RNNs. In our approach, dropout is carefully used in the network so that it does not affect the recurrent connections, hence the power of RNNs in modeling sequences is preserved. Extensive experiments on a broad range of handwritten databases confirm the effectiveness of dropout on deep architectures even when the network mainly consists of recurrent and shared connections.

/pdf/dropout-improves-recurrent-neural-networks-for-handwriting-2di6lmbs26.pdf

Dropout Improves Recurrent Neural Networks for Handwriting Recognition

This paper describes the NIST 2013 Open Handwriting Recognition and Translation evaluation (OpenHaRT'13). A short background leading to the start of OpenHaRT is included. The test designs pertaining to the tasks, the data used, the performance measurements, and the protocols are presented. The participants and their submissions are mentioned followed by the evaluation results and some preliminary analyses. The paper concludes with some thoughts toward future evaluations.

NIST 2013 Open Handwriting Recognition and Translation (Open HaRT'13) Evaluation

This competition is aimed at classification of writer demographics from offline handwritten documents using the QUWI database. QUWI is a bilingual database comprising writing samples of same individuals in Arabic and English. This allows evaluating the performance of different systems in a more challenging multi-script environment. This paper presents the details of the competition tasks, the datasets used in each of the tasks, a brief description of the participating systems, experimental protocol and evaluation criteria and finally the overall rankings of the participants.

ICFHR2016 Competition on Multi-script Writer Demographics Classification Using "QUWI" Database

This paper presents a real database for the Arabic printed text recognition, APTID / MF (Arabic Printed Text Image Database / Multi-Font).This database can be used to evaluate the system that recognizes Arabic printed texts with an open vocabulary. APTID / MF may be also used for research in word segmentation and font identification. APTID / MF is obtained from 387 pages of Arabic printed documents scanned with grayscale format and 300 dpi resolutions. From this documents, 1,845 text-blocks have been extracted. In addition ground truth file is provided for each texts-block. APTID / MF also includes an Arabic printed character image dataset made up of 27,402 samples. The database is freely available to interested researchers.

/pdf/database-for-arabic-printed-text-recognition-research-3qy6hfmtvy.pdf

Database for Arabic Printed Text Recognition Research

Writer identification still remains as a challenge area in the field of off-line handwriting recognition because only an image of the handwriting is available. Consequently, some information on the dynamic of writing, which is valuable for identification of writer, is unavailable in the off-line approaches, contrary to the on-line approaches where temporal and spatial information for the handwriting is available. In this paper we present a new method for writer identification based on Multi-Fractal features for both types of presented approaches. This method consists to extract the multi-fractal dimensions from the images of Arabic words and the on-line signals for the same words. In order to enhance the performance of our writer identification system, we have combined both on-line and off-line approaches, taking the advantage it provides ADAB database, which allows to recover the on-line signal and image for the same handwriting. In this way, our work consists to take advantage of static and dynamic representations of handwriting, in order to identify the writer in realistic conditions. The tests are performed on the writing of 100 writers from the ADAB database. The obtained results show the effectiveness of the proposed writer identification system.

Combining of Off-line and On-line Feature Extraction Approaches for Writer Identification

In this paper, we present a comparison between two different combination schemes for the improvement of the performance of Arabic handwriting recognition systems. Several recognition systems (here considered as black box systems) are used from the participating systems of the Arabic handwriting recognition competition at ICDAR 2007. The outputs of these systems provide the input of our combination schemes. The first combination schemes are based on fixed fusion using logical rules, while the second one are based on trainable rules. After the normalization step of the recognition confidences and the combination of the outputs, the improvement is evaluated in term of recognition rates of a multi-classifier system with or without reject. The participating systems use the sets a to e of the IfN/ENIT database for training, and we use the set f for tests. Applying the combination rules, the results show a high recognition rate of about 95% without reject, which corresponds to an improvement of recognition rates between 8% and 15% compared to results at the ICDAR 2007 competition.

Haikal El Abed

Papers

NIST 2013 Open Handwriting Recognition and Translation (Open HaRT'13) Evaluation

ICFHR2016 Competition on Multi-script Writer Demographics Classification Using "QUWI" Database

Database for Arabic Printed Text Recognition Research

Combining of Off-line and On-line Feature Extraction Approaches for Writer Identification

Improvement of Arabic Handwriting Recognition Systems - Combination and/or Reject?