Offline Arabic Handwriting Recognition with Multidimensional Recurrent Neural Networks

Graves, Alex; Schmidhuber, Jürgen

doi:10.1007/978-1-4471-4072-6_12

Cited by 533 publications

(584 citation statements)

References 8 publications

Supporting

Mentioning

573

Contrasting

Unclassified

Order By: Relevance

“…In the experi -ments, each single image was divided into [3] 95. 45 94.48 SDIP [5] 93.00 91.60 MFA [34] 91.60 90.10 regions, and a maximum of 13 regions were randomly dropped. The results are shown in Table 4.…”

Section: Evaluation Of Dropregion In Comparison To Dropoutmentioning

confidence: 99%

DropRegion training of inception font network for high-performance Chinese font recognition

Huang

Zhong

Jin

et al. 2018

Pattern Recognition

View full text Add to dashboard Cite

Chinese font recognition (CFR) has gained significant attention in recent years. However, due to the sparsity of labeled font samples and the structural complexity of Chinese characters, CFR is still a challenging task. In this paper, a DropRegion method is proposed to generate a large number of stochastic variant font samples whose local regions are selectively disrupted and an inception font network (IFN) with two additional convolutional neural network (CNN) structure elements, i.e., a cascaded cross-channel parametric pooling (CCCP) and global average pooling, is designed. Because the distribution of strokes in a font image is non-stationary, an elastic meshing technique that adaptively constructs a set of local regions with equalized information is developed. Thus, DropRegion is seamlessly embedded in the IFN, which enables end-to-end training; the proposed DropRegion -IFN can be used for high performance CFR. Experimental results have confirmed the effectiveness of our new approach for CFR.

show abstract

Section: Evaluation Of Dropregion In Comparison To Dropoutmentioning

confidence: 99%

DropRegion training of inception font network for high-performance Chinese font recognition

Huang

Zhong

Jin

et al. 2018

Pattern Recognition

View full text Add to dashboard Cite

show abstract

“…To do this I utilized two R libraries, nnet 155 and neuralnet 156 . I found that the two libraries under default parameters (set to identify categorical not continuous variables) produce 152 (Farley and Clark, 1954) 153 (Graves and Schmidhuber, 2009) 154 The machine used was a server with 16 CPU cores, and 128gb of RAM. 155 (Ripley and Venables, 2011) 156 (Fritsch, Guenther and Guenther, 2012) consistently different results.…”

Section: Modeling Approach Comparisonmentioning

confidence: 99%

Tweeting About Mental Health: Big Data Text Analysis of Twitter for Public Policy

Zaydman¹

2017

View full text Add to dashboard Cite

“…Multi Convolution Neural Networks are applied in Graves et al (2008). Combination of HMM and Time delay Neural Networks have shown good performance for cursive script recognition in Madhvanath et al (2007).…”

Section: Neural Network Modelsmentioning

confidence: 99%

Preprocessing of online handwritten Telugu character recognition

Inuganti¹,

Ramisetty

2017

Int. j. adv. appl. sci.

View full text Add to dashboard Cite

Online Handwritten Character Recognition (OHCR) is the method of recognizing characters by a machine while the user writes, in which the handheld devices record (x, y) coordinates of the track of the character. With the advent of handheld devices, there is a great attention towards OHCR of regional languages. Preprocessing is the main phase, in OHCR, as it increases the performance of succeeding phases, by removing the inconsistency or the redundancy present in the data collected in real-world environment. In this paper, we depict the model of Preprocessing of Online Handwritten Telugu Strokes. The preprocessing steps we address in our article are Normalization, Smoothing, Duplicate Point Removal, Interpolation, Dehooking and Resampling. Preprocessing data performance is evaluated through parameters namely recognition accuracy, recognition speed, false acceptance rate and false rejection rate over HP labs dataset hpl-Telugu-ISO-char-online-1.0. The dataset contains samples of the 166 character classes collected of different writers on ACECAD Digimemo (A4 sized) using an AcecadDigi memo DCT application. It consists of 270 samples on average for each of 166 Telugu "characters" written by native Telugu writers.

show abstract

Offline Arabic Handwriting Recognition with Multidimensional Recurrent Neural Networks

Cited by 533 publications

References 8 publications

DropRegion training of inception font network for high-performance Chinese font recognition

DropRegion training of inception font network for high-performance Chinese font recognition

Tweeting About Mental Health: Big Data Text Analysis of Twitter for Public Policy

Preprocessing of online handwritten Telugu character recognition

Contact Info

Product

Resources

About