Breast Cancer Classification from Ultrasound Images Using Probability-Based Optimal Deep Learning Feature Fusion

Sensors, Journal Year: 2022, Volume and Issue: 22(3), P. 807 - 807

Published: Jan. 21, 2022

After lung cancer, breast cancer is the second leading cause of death in women. If detected early, mortality rates women can be reduced. Because manual diagnosis takes a long time, an automated system required for early detection. This paper proposes new framework classification from ultrasound images that employs deep learning and fusion best selected features. The proposed divided into five major steps: (i) data augmentation performed to increase size original dataset better Convolutional Neural Network (CNN) models; (ii) pre-trained DarkNet-53 model considered output layer modified based on augmented classes; (iii) trained using transfer features are extracted global average pooling layer; (iv) two improved optimization algorithms known as reformed differential evaluation (RDE) gray wolf (RGW); (v) fused probability-based serial approach classified machine algorithms. experiment was conducted Breast Ultrasound Images (BUSI) dataset, accuracy 99.1%. When compared with recent techniques, outperforms them.

Language: Английский

---

Recognition of Leaf Disease Using Hybrid Convolutional Neural Network by Applying Feature Reduction

Sensors, Journal Year: 2022, Volume and Issue: 22(2), P. 575 - 575

Published: Jan. 12, 2022

Agriculture is crucial to the economic prosperity and development of India. Plant diseases can have a devastating influence towards food safety considerable loss in production agricultural products. Disease identification on plant essential for long-term agriculture sustainability. Manually monitoring difficult due time limitations diversity diseases. In realm inputs, automatic characterization widely required. Based performance out all image-processing methods, better suited solving this task. This work investigates grapevines. Leaf blight, Black rot, stable, measles are four types found grape plants. Several earlier research proposals using machine learning algorithms were created detect one or two leaves; no offers complete detection The photos taken from village dataset order use transfer retrain EfficientNet B7 deep architecture. Following learning, collected features down-sampled Logistic Regression technique. Finally, most discriminant traits identified with highest constant accuracy 98.7% state-of-the-art classifiers after 92 epochs. simulation findings, an appropriate classifier application also suggested. proposed technique’s effectiveness confirmed by fair comparison existing procedures.

Language: Английский

---

Image Denoising Using a Compressive Sensing Approach Based on Regularization Constraints

Sensors, Journal Year: 2022, Volume and Issue: 22(6), P. 2199 - 2199

Published: March 11, 2022

In remote sensing applications and medical imaging, one of the key points is acquisition, real-time preprocessing storage information. Due to large amount information present in form images or videos, compression these data necessary. Compressed an efficient technique meet this challenge. It consists acquiring a signal, assuming that it can have sparse representation, by using minimum number nonadaptive linear measurements. After compressed process, reconstruction original signal must be performed at receiver. Reconstruction techniques are often unable preserve texture image tend smooth out its details. To overcome problem, we propose, work, method combines total variation regularization non-local self-similarity constraint. The optimization augmented Lagrangian avoids difficult problem nonlinearity nondifferentiability terms. proposed algorithm, called denoising-compressed (DCSR) terms, will not only perform but also denoising. evaluate performance compare with state-of-the-art methods, such as Nesterov's group-based representation wavelet-based terms denoising preservation edges, details, well from point view computational complexity. Our approach permits gain up 25% efficiency visual quality two metrics: peak signal-to-noise ratio (PSNR) structural similarity (SSIM).

Language: Английский

---

Grade Prediction Modeling in Hybrid Learning Environments for Sustainable Engineering Education

Sustainability, Journal Year: 2022, Volume and Issue: 14(9), P. 5205 - 5205

Published: April 26, 2022

Since mid-March 2020, due to the COVID-19 pandemic, higher education has been facing a very uncertain situation, despite hasty implementation of information and communication technologies for distance online learning. Hybrid learning, i.e., mixing face-to-face seems be rule in most universities today. In order build post-COVID-19 university education, one that is increasingly digital sustainable, it essential learn from these years health crisis. this context, paper aims identify quantify main factors affecting mechanical engineering student performance generalized linear autoregressive (GLAR) model. This model, which distinguished by its simplicity ease implementation, responsible predicting grades learning situations hybrid environments. The thirty or so variables identified previously tested model 2020–2021, was exclusive mode were evaluated blended spaces. Given low predictive power original about ten new factors, specific then tested. refined version GLAR predicts within ±1 with success rate 63.70%, making 28.08% more accurate than originally created 2020–2021. Special attention also given students whose grade predictions underestimated who failed. methodology presented applicable all aspects academic process, including students, instructors, decisionmakers.

Language: Английский

---

Advanced genetic image encryption algorithms for intelligent transport systems

Computers & Electrical Engineering, Journal Year: 2025, Volume and Issue: 123, P. 110162 - 110162

Published: Feb. 17, 2025

---

Concrete Cracks Detection and Monitoring Using Deep Learning-Based Multiresolution Analysis

Electronics, Journal Year: 2021, Volume and Issue: 10(15), P. 1772 - 1772

Published: July 24, 2021

In this paper, we propose a new methodology for crack detection and monitoring in concrete structures. This approach is based on multiresolution analysis of sample or specimen material subjected to several types solicitation. The image obtained by ultrasonic investigation processed customized wavelet analyzed at various scales order detect internal cracks initiation. ultimate objective work an automatic type identification scheme convolutional neural networks (CNN). context, propagation can be monitored without access the surface goal before they are visible. achieved through combination two major data tools which wavelets deep learning. original procedure shown yield high accuracy close 90%. evaluate performance proposed CNN architectures, also used open database, SDNET2018, external cracks.

Language: Английский

---

Eye Aspect Ratio for Real-Time Drowsiness Detection to Improve Driver Safety

Electronics, Journal Year: 2022, Volume and Issue: 11(19), P. 3183 - 3183

Published: Oct. 4, 2022

Drowsiness is a major risk factor for road safety, contributing to serious injury, death, and economic loss on the road. Driving performance decreases because of increased drowsiness. In several different applications, such as facial movement analysis driver blink detection an essential requirement that used. The extremely rapid rate, other hand, makes automatic challenging task. This research paper presents technique identifying eye blinks in video series recorded by car dashboard camera real time. suggested determines landmark positions each frame then extracts vertical distance between eyelids from positions. algorithm has been proposed estimates positions, single scalar quantity making use Eye Aspect Ratio (EAR), identifies closeness frame. end, are recognized employing modified EAR threshold value conjunction with pattern values relatively short period Experimental evidence indicates greater threshold, worse AUC’s accuracy performance. Further, 0.18 was determined be optimum our research.

Language: Английский

---

AuSR1: Authentication and self-recovery using a new image inpainting technique with LSB shifting in fragile image watermarking

Journal of King Saud University - Computer and Information Sciences, Journal Year: 2022, Volume and Issue: 34(8), P. 5822 - 5840

Published: Feb. 25, 2022

Language: Английский

---

A Comprehensive survey on ear recognition: Databases, approaches, comparative analysis, and open challenges

Neurocomputing, Journal Year: 2023, Volume and Issue: 537, P. 236 - 270

Published: March 30, 2023

Language: Английский

---

Towards Explainable Ear Recognition Systems Using Deep Residual Networks

IEEE Access, Journal Year: 2021, Volume and Issue: 9, P. 122254 - 122273

Published: Jan. 1, 2021

This paper presents ear recognition models constructed with Deep Residual Networks (ResNet) of various depths. Due to relatively limited amounts images we propose three different transfer learning strategies address the problem. is done either through utilizing ResNet architectures as feature extractors or employing end-to-end system designs. First, use pretrained trained on specific visual tasks, inititalize network weights and train fully-connected layer task. Second, fine-tune entire training part each dataset. Third, utilize output penultimate fine-tuned feed SVM classifiers. Finally, build ensembles networks depths enhance overall performance. Extensive experiments are conducted evaluate obtained using acquired under constrained unconstrained imaging conditions from AMI, AMIC, WPUT AWE databases. The best performance by averaging achieving accuracy 99.64%, 98.57%, 81.89%, 67.25% WPUT, databases, respectively. In order facilitate interpretation results explain differences dataset apply powerful Guided Grad-CAM technique, which provides explanations unravel black-box nature deep models. provided visualizations highlight most relevant discriminative regions exploited differentiate between individuals. Based our analysis localization maps argue that make correct prediction when considering geometrical structure shape a region even mild degree head rotations presence hair occlusion accessories. However, severe movements low contrast have negative impact

Language: Английский

---