A new Arabic handwritten character recognition deep learning system (AHCR-DLS)

The sudden appearance of COVID-19 has put the world in a serious situation. Due to the rapid spread of the virus and the increase in the number of infected patients and deaths, COVID-19 was declared a pandemic. This pandemic has its destructive effect not only on humans but also on the economy. Despite the development and availability of different vaccines for COVID-19, scientists still warn the citizens of new severe waves of the virus, and as a result, fast diagnosis of COVID-19 is a critical issue. Chest imaging proved to be a powerful tool in the early detection of COVID-19. This study introduces an entire framework for the early detection and early prognosis of COVID-19 severity in the diagnosed patients using laboratory test results. It consists of two phases (1) Early Diagnostic Phase (EDP) and (2) Early Prognostic Phase (EPP). In EDP, COVID-19 patients are diagnosed using CT chest images. In the current study, 5, 159 COVID-19 and 10, 376 normal computed tomography (CT) images of Egyptians were used as a dataset to train 7 different convolutional neural networks using transfer learning. Data augmentation normal techniques and generative adversarial networks (GANs), CycleGAN and CCGAN, were used to increase the images in the dataset to avoid overfitting issues. 28 experiments were applied and multiple performance metrics were captured. Classification with no augmentation yielded accuracy by EfficientNetB7 architecture. By applying CycleGAN and CC-GAN Augmentation, the maximum reported accuracies were and by MobileNetV1 and VGG-16 architectures respectively. In EPP, the prognosis of the severity of COVID-19 in patients is early determined using laboratory test results. In this study, 25 different classification techniques were applied and from the different results, the highest accuracies were and reported by the Ensemble Bagged Trees and Tree (Fine, Medium, and Coarse) techniques respectively.

show abstract

“…The output of the pooling layer is then flattened and transformed into a vector of numbers (Balaha et al. 2021d ).…”

Section: Convolutional Neural Network (Cnns)mentioning

confidence: 99%

A complete framework for accurate recognition and prognosis of COVID-19 patients based on deep transfer learning and feature classification approach

2022

Self Cite

View full text Add to dashboard Cite

show abstract

“…This data was fed to a CNN model which gave an accuracy of 94.9%. Recent works on applying CNN-based deep learning approaches have resulted in better accuracy ( Balaha et al, 2021a ; Balaha et al, 2021b ). Deep learning has also been combined with other approaches such as Mathematical Morphology Operations (MMO) to provide for better character recognition ( Elkhayati, Elkettani & Mourchid, 2022 ).…”

Section: Related Workmentioning

confidence: 99%

Writer verification of partially damaged handwritten Arabic documents based on individual character shapes

Khan

Mohammad

Brahim

et al. 2022

PeerJ Computer Science

View full text Add to dashboard Cite

Author verification of handwritten text is required in several application domains and has drawn a lot of attention within the research community due to its importance. Though, several approaches have been proposed for the text-independent writer verification of handwritten text, none of these have addressed the problem domain where author verification is sought based on partially-damaged handwritten documents (e.g., during forensic analysis). In this paper, we propose an approach for offline text-independent writer verification of handwritten Arabic text based on individual character shapes (within the Arabic alphabet). The proposed approach enables writer verification for partially damaged documents where certain handwritten characters can still be extracted from the damaged document. We also provide a mechanism to identify which Arabic characters are more effective during the writer verification process. We have collected a new dataset, Arabic Handwritten Alphabet, Words and Paragraphs Per User (AHAWP), for this purpose in a classroom setting with 82 different users. The dataset consists of 53,199 user-written isolated Arabic characters, 8,144 Arabic words, 10,780 characters extracted from these words. Convolutional neural network (CNN) based models are developed for verification of writers based on individual characters with an accuracy of 94% for isolated character shapes and 90% for extracted character shapes. Our proposed approach provided up to 95% writer verification accuracy for partially damaged documents.

show abstract

“…F1-score is the harmonic mean of precision and recall as shown in Eq. (27) ( Balaha, Ali, Saraya, & Badawy, 2021 ).…”

Section: Covh2sd: a Hybrid Harris Hawks Optimization Deep Learning Approachmentioning

confidence: 99%

CovH2SD: A COVID-19 detection approach based on Harris Hawks Optimization and stacked deep learning

Balaha

El-Gendy

Saafan

2021

Expert Systems with Applications

Self Cite

View full text Add to dashboard Cite

Starting from Wuhan in China at the end of 2019, coronavirus disease (COVID-19) has propagated fast all over the world, affecting the lives of billions of people and increasing the mortality rate worldwide in few months. The golden treatment against the invasive spread of COVID-19 is done by identifying and isolating the infected patients, and as a result, fast diagnosis of COVID-19 is a critical issue. The common laboratory test for confirming the infection of COVID-19 is Reverse Transcription Polymerase Chain Reaction (RT-PCR). However, these tests suffer from some problems in time, accuracy, and availability. Chest images have proven to be a powerful tool in the early detection of COVID-19. In the current study, a hybrid learning and optimization approach named CovH2SD is proposed for the COVID-19 detection from the Chest Computed Tomography (CT) images. CovH2SD uses deep learning and pre-trained models to extract the features from the CT images and learn from them. It uses Harris Hawks Optimization (HHO) algorithm to optimize the hyperparameters. Transfer learning is applied using nine pre-trained convolutional neural networks (i.e. ResNet50, ResNet101, VGG16, VGG19, Xception, MobileNetV1, MobileNetV2, DenseNet121, and DenseNet169). Fast Classification Stage (FCS) and Compact Stacking Stage (CSS) are suggested to stack the best models into a single one. Nine experiments are applied and results are reported based on the Loss, Accuracy, Precision, Recall, F1-Score, and Area Under Curve (AUC) performance metrics. The comparison between combinations is applied using the Weighted Sum Method (WSM). Six experiments report a WSM value above 96.5%. The top WSM and accuracy reported values are 99.31% and 99.33% respectively which are higher than the eleven compared state-of-the-art studies

show abstract

A new Arabic handwritten character recognition deep learning system (AHCR-DLS)

Cited by 69 publications

References 40 publications

A complete framework for accurate recognition and prognosis of COVID-19 patients based on deep transfer learning and feature classification approach

A complete framework for accurate recognition and prognosis of COVID-19 patients based on deep transfer learning and feature classification approach

Writer verification of partially damaged handwritten Arabic documents based on individual character shapes

CovH2SD: A COVID-19 detection approach based on Harris Hawks Optimization and stacked deep learning

Contact Info

Product

Resources

About