Deep Learning Network Architecture based Kannada Handwritten Character Recognition

Deep metric learning is one of the recommended methods for the challenge of supporting few/zero-shot learning by deep networks. It depends on building a Siamese architecture of two homogeneous Convolutional Neural Networks (CNNs) for learning a distance function that can map input data from the input space to the feature space. Instead of determining the class of each sample, the Siamese architecture deals with the existence of a few training samples by deciding if the samples share the same class identity or not. The traditional structure for the Siamese architecture was built by forming two CNNs from scratch with randomly initialized weights and trained by binary cross-entropy loss. Building two CNNs from scratch is a trial and error and time-consuming phase. In addition, training with binary crossentropy loss sometimes leads to poor margins. In this paper, a novel Siamese network is proposed and applied to few/zero-shot Handwritten Character Recognition (HCR) tasks. The novelties of the proposed network are in. 1) Utilizing transfer learning and using the pre-trained AlexNet as a feature extractor in the Siamese architecture. Fine-tuning a pre-trained network is typically faster and easier than building from scratch. 2) Training the Siamese architecture with contrastive loss instead of the binary cross-entropy. Contrastive loss helps the network to learn a nonlinear mapping function that enables it to map the extracted features in the vector space with an optimal way. The proposed network is evaluated on the challenging Chars74K datasets by conducting two experiments. One is for testing the proposed network in few-shot learning while the other is for testing it in zero-shot learning. The recognition accuracy of the proposed network reaches to 85.6% and 82% in few-and zero-shot learning respectively. In addition, a comparison between the performance of the proposed Siamese network and the traditional Siamese CNNs is conducted. The comparison results show that the proposed network achieves higher recognition results in less time. The proposed network reduces the training time from days to hours in both experiments.

show abstract

“…CNN [39] 57% Transfer learning from Devanagari handwritten recognition system + VGG19 NET [40] 73.51%…”

Section: Modelmentioning

confidence: 99%

A Novel Siamese Network for Few/Zero-Shot Handwritten Character Recognition Tasks

Elaraby¹,

Barakat²,

Rezk³

2023

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

show abstract

“…Rani, N. Shobha., (2020) [24] suggested that the handwritten Alyoubi et al, (2020) [25] planned that Diabetic Retinopathy (DR) causes abnormalities in the retina that impede vision when they're not well planned. Due to the irreversibility of DR, therapy can only sustain eyesight, not restore it.…”

Section: Literature Surveymentioning

confidence: 99%

“…It is the most widely used technique for recognizing patterns. CNN Technique Accuracy (%) OCR [24] 99% OCR [32] 97% DLN [25] 73.51% DR [27] 94.5% Accuracy (%) OCR [27] OCR [28] DLN [32] DR [35]…”

Section: Comparative Analysismentioning

confidence: 99%

A review of english language from the scanned document with the use of pattern recognition: Detailed study on methodology and working of optical recognition system

Moolchandani,

Hundal

2022

Int. J. Comput. Programming Database Manage.

View full text Add to dashboard Cite

Optical Character Recognition (OCR) is the procedure through which a machine analyses scanned printed or handwritten documents. OCR is a new study topic for Indian scripts that has the potential to replace the habit of manually completing data entries for diverse documents from printed paper records. The primary benefit of this digital textual material is that it can be altered, something that is not feasible with digitized files.

show abstract

“…Rani, N. Shobha., (2020) [24] suggested that the handwritten Kannada character identification based on Devanagari handwriting recognition technology can be used to impart knowledge. The enormous Devanagari recognition system data corpus will be used as training material to recognize traditional Kannada characters written by hand, albeit with a smaller database.VGG19 NET uses deep learning network architecture to transfer knowledge for recognition.…”

Section: Literature Surveymentioning

confidence: 99%

A review of English language from the scanned document with the use of pattern recognition: Detailed study on recognition system

Moolchandani,

Hundal

2022

Int. J. Comput. Artif. Intell.

View full text Add to dashboard Cite

The technology of Document Analysis and Recognition, as a branch of pattern recognition, faces various practical demands in the real world, such as the digitization of books, newspapers, and archives, invoicing, and corporate documents. Pattern recognition and machine learning are among the most cutting-edge areas in software science. Statistical learning theory based neural network approaches and methodologies have recently come under more and more scrutiny. By utilizing the proper character recognition and segmentation modules for optical character recognition (OCR), it is essential to identify the document's language and printing style. In the field of pattern recognition, recognizing handwritten papers is a difficult problem. Algorithms and statistical models that computers use to complete a certain task without being explicitly focused on machine learning (ML). It is possible to utilize these algorithms for a variety of purposes, including data mining and image processing. It is easy to automate tasks by utilizing machine learning when an algorithm has learned how to deal with data. The purpose of this review article is to summarize and compare many well-known methodologies that are utilized at various phases of a pattern recognition system's development. After examining several strategies for pattern identification, it was determined that the most accurate method is optical character recognition (OCR). Optical character recognition (OCR) scanners, on the other hand, have a 99 percent accuracy rate. Diabetes Retinopathy (DR) also has the lowest accuracy, at just percent.

show abstract

Deep Learning Network Architecture based Kannada Handwritten Character Recognition

Cited by 39 publications

References 7 publications

A Novel Siamese Network for Few/Zero-Shot Handwritten Character Recognition Tasks

A Novel Siamese Network for Few/Zero-Shot Handwritten Character Recognition Tasks

A review of english language from the scanned document with the use of pattern recognition: Detailed study on methodology and working of optical recognition system

A review of English language from the scanned document with the use of pattern recognition: Detailed study on recognition system

Contact Info

Product

Resources

About