Recognizing handwritten single digits and digit strings using deep architecture of neural networks

Saabni, Raid

doi:10.1109/icaipr.2016.7585206

Cited by 25 publications

(10 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, considerable efforts have been made to employ deep learning techniques within HDSR, especially the Convolutional Neural Networks (CNNs) and its advanced automatic feature extractor [31]. In this way, Saabni [5] proposed an algorithm that trains k-Sparse Auto Encoders and stacks its hidden layers as pre-trained hidden layers within a deep neural network, in which he used sliding window technique to achieve high recognition rates.…”

Section: B State-of-the-art In Hdsrmentioning

confidence: 99%

“…In addition, the very low volume datasets for training optical models represent a great challenge for deep learning techniques, in which the vanishing gradient problem increases according to the complexity of the model (how deep) [14]. Thus, several works were developed to solve this problem and achieve better results, such as models composed by K-sparse Auto Encoders [5], CRNN [15], ResNet with RNN-CTC [6], [7], CNN [16], ResNet-41 [17], and YoLo [8].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

HDSR-Flor: A Robust End-to-End System to Solve the Handwritten Digit String Recognition Problem in Real Complex Scenarios

et al. 2020

View full text Add to dashboard Cite

Automatic handwriting recognition systems are of interest for academic research fields and for commercial applications. Recent advances in deep learning techniques have shown dramatic improvement in relation to classic computer vision problems, especially in Handwritten Text Recognition (HTR). However, several approaches try to solve the problem of deep learning applied to Handwritten Digit String Recognition (HDSR), where it has to deal with the low number of trainable data, while learning to ignore any writing symbol around the digits (noise). In this context, we present a new optical model architecture (Gated-CNN-BGRU), based on HTR workflow, applied to HDSR. The International Conference on Frontiers of Handwriting Recognition (ICFHR) 2014 competition on HDSR were used as baselines to evaluate the effectiveness of our proposal, whose metrics, datasets and recognition methods were adopted for fair comparison. Furthermore, we also use a private dataset (Brazilian Bank Check-Courtesy Amount Recognition), and 11 different approaches from the state-of-the-art in HDSR, as well as 2 optical models from the state-of-the-art in HTR. Finally, the proposed optical model demonstrated robustness even with low data volume (126 trainable data, for example), surpassing the results of existing methods with an average precision of 96.50%, which is equivalent to an average percentage of improvement of 3.74 points compared to the state-of-the-art in HDSR. In addition, the result stands out in the competition's CVL HDS set, where the proposed optical model achieved a precision of 93.54%, while the best result so far had been from Beijing group (from the competition itself), with 85.29%.

show abstract

Section: B State-of-the-art In Hdsrmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

HDSR-Flor: A Robust End-to-End System to Solve the Handwritten Digit String Recognition Problem in Real Complex Scenarios

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The problem is that we do not usually know the number of digits in the string and so the optimal boundary between them is unknown. Such a problem has been dealt with in different ways [5], [9], [14] and one way to approach it is to see it as a sequence-to-sequence problem with a sequence of image patches as input and a sequence of characters as output [2]. To solve sequenceto-sequence problems with different length from input to output, the key point is to find the alignment between them.…”

Section: Introductionmentioning

confidence: 99%

On the Use of Attention Mechanism in a Seq2Seq Based Approach for Off-Line Handwritten Digit String Recognition

Lupinski

Belaïd

Echi

2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

In this work, we investigate the use of the attention mechanism in deep learning for a better reading of handwritten digit strings in digitized images. The proposed recognition system built upon a CNN (Convolutional Neural Network) and two RNNs (Recurrent Neural Networks), acting as Encoder and Decoder and using the attention mechanism. We used a 1D mechanism for attention location with a "soft" alignment attention which has the peculiarity of having an easily calculable gradient and thus to integrate well with the network. Experimental results on data from ORAND-CAR A, ORAND-CAR B and CVL HDS databases compare favorably to other published methods.

show abstract

“…Recent years, with the advancement of deep learning, handwritten digit string recognition (HDSR) has archived great improvements [1][2][3]. An intuitive approach to recognize these handwriting strings is to segment string images into pieces which correspond to single characters or part of them, then combine the recognition results of these pieces with path-search algorithms to get global optimal results.…”

Section: Introductionmentioning

confidence: 99%

“…This method won the first place on the ICFHR2014 HDSR competition [1]. Saabni [3] used sliding window and deep neural network to attain high recognition rates. Gattal et al [2] applied three segmentation methods to handle handwritten digit strings by combining these segmentation methods depending on the configuration link between digits.…”

Section: Introductionmentioning

confidence: 99%

Handwritten Digit String Recognition by Combination of Residual Network and RNN-CTC

Wang

2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Recurrent neural network (RNN) and connectionist temporal classification (CTC) have showed successes in many sequence labeling tasks with the strong ability of dealing with the problems where the alignment between the inputs and the target labels is unknown. Residual network is a new structure of convolutional neural network and works well in various computer vision tasks. In this paper, we take advantage of the architectures mentioned above to create a new network for handwritten digit string recognition. First we design a residual network to extract features from input images, then we employ a RNN to model the contextual information within feature sequences and predict recognition results. At the top of this network, a standard CTC is applied to calculate the loss and yield the final results. These three parts compose an end-to-end trainable network. The proposed new architecture achieves the highest performances on ORAND-CAR-A and ORAND-CAR-B with recognition rates 89.75% and 91.14%, respectively. In addition, the experiments on a generated captcha dataset which has much longer string length show the potential of the proposed network to handle long strings.

show abstract

Recognizing handwritten single digits and digit strings using deep architecture of neural networks

Cited by 25 publications

References 11 publications

HDSR-Flor: A Robust End-to-End System to Solve the Handwritten Digit String Recognition Problem in Real Complex Scenarios

HDSR-Flor: A Robust End-to-End System to Solve the Handwritten Digit String Recognition Problem in Real Complex Scenarios

On the Use of Attention Mechanism in a Seq2Seq Based Approach for Off-Line Handwritten Digit String Recognition

Handwritten Digit String Recognition by Combination of Residual Network and RNN-CTC

Contact Info

Product

Resources

About