Wyniki wyszukiwania - BazTech

1

Remaining useful life prediction of a lithium-ion battery based on a temporal convolutional network with data extension

Zhao Jing, Liu Dayong, Meng Lingshuai

International Journal of Applied Mathematics and Computer Science

|

2024

|

Vol. 34, no. 1

105--117

EN

Unmanned underwater vehicles are typically deployed in deep sea environments, which present unique working conditions. Lithium-ion power batteries are crucial for powering underwater vehicles, and it is vital to accurately predict their remaining useful life (RUL) to maintain system reliability and safety. We propose a residual life prediction model framework based on complete ensemble empirical mode decomposition with an adaptive noise-temporal convolutional net (CEEMDAN-TCN), which utilizes dilated causal convolutions to improve the model’s ability to capture local capacity regeneration and enhance the overall prediction accuracy. CEEMDAN is employed to denoise the data and prevent RUL prediction errors caused by local regeneration, and feature expansion is utilized to extend the temporal dimension of the original data. The NASA and CALCE battery capacity datasets are used as input to train the network framework. The output is the current predicted residual capacity, which is compared with the real residual battery capacity. The MAE, RMSE and RE are used as the evaluation indexes of the RUL prediction performance. The proposed network model is verified on the NASA and CACLE datasets. The evaluation results show that our method has better life prediction performance. At the same time, it is proved that both feature expansion and modal decomposition can improve the generalization ability of the model, which is very useful in industrial scenarios.

2

A hybrid lightweight breast cancer classification framework using the histopathological images

Addo Daniel, Zhou Shijie, Sarpong Kwabena, Nartey Obed T., Abdullah Muhammed A., Ukwuoma Chiagoziem C., Al-antari Mugahed A.

Biocybernetics and Biomedical Engineering

|

2024

|

Vol. 44, no. 1

31--54

EN

A crucial element in the diagnosis of breast cancer is the utilization of a classification method that is efficient, lightweight, and precise. Convolutional neural networks (CNNs) have garnered attention as a viable approach for classifying histopathological images. However, deeper and wider models tend to rely on first-order statistics, demanding substantial computational resources and struggling with fixed kernel dimensions that limit encompassing diverse resolution data, thereby degrading the model’s performance during testing. This study introduces BCHI-CovNet, a novel lightweight artificial intelligence (AI) model for histopathological breast image classification. Firstly, a novel multiscale depth-wise separable convolution is proposed. It is introduced to split input tensors into distinct tensor fragments, each subject to unique kernel sizes integrating various kernel sizes within one depth-wise convolution to capture both low- and high-resolution patterns. Secondly, an additional pooling module is introduced to capture extensive second-order statistical information across the channels and spatial dimensions. This module works in tandem with an innovative multi-head self-attention mechanism to capture the long-range pixels contributing significantly to the learning process, yielding distinctive and discriminative features that further enrich representation and introduce pixel diversity during training. These novel designs substantially reduce computational complexities regarding model parameters and FLOPs, which is crucial for resource-constrained medical devices. The outcomes achieved by employing the suggested model on two openly accessible datasets for breast cancer histopathological images reveal noteworthy performance. Specifically, the proposed approach attains high levels of accuracy: 99.15 % at 40× magnification, 99.08 % at 100× magnification, 99.22 % at 200× magnification, and 98.87 % at 400× magnification on the BreaKHis dataset. Additionally, it achieves an accuracy of 99.38 % on the BACH dataset. These results highlight the exceptional effectiveness and practical promise of BCHI-CovNet for the classification of breast cancer histopathological images.

3

Deep learning correction for image reconstruction in electrical impedance tomography using UNet model

Maciura Łukasz, Rymarczyk Tomasz, Wójcik Dariusz, Gauda Konrad, Kowalski Marcin

Przegląd Elektrotechniczny

|

2023

|

R. 99, nr 6

134--137

EN

This article was inspired by a similar Deep DBar algorithm, where a modified UNet convolutional model was used to correct the output of the DBar algorithm using the UNet model. However, instead of the DBar algorithm, another deterministic electrical impedance tomography reconstruction algorithm was used in this solution. The modified UNet model was used to successfully correct the initial reconstructions, which were computed using Kotre regularities using pseudo-inversion of the sensitivity matrix.

PL

Ten artykuł został inspirowany podobnym algorytmem Deep DBar, w którym zmodyfikowany model splotowy UNet został użyty do skorygowania danych wyjściowych algorytmu DBar przy użyciu modelu UNet. Jednak zamiast algorytmu DBar w tym rozwiązaniu zastosowano inny deterministyczny algorytm rekonstrukcji elektrycznej tomografii impedancyjnej. Zmodyfikowany model UNet został wykorzystany do skutecznej korekcji wstępnych rekonstrukcji, które zostały obliczone przy użyciu regularności Kotrego z wykorzystaniem pseudo-inwersji macierzy czułości.

4

MDCF_Net: A Multi-dimensional hybrid network for liver and tumor segmentation from CT

Jiang Jian, Peng Yanjun, Hou Qingfan, Wang Jiao

Biocybernetics and Biomedical Engineering

|

2023

|

Vol. 43, no. 2

494--506

EN

The segmentation of the liver and liver tumors is critical in the diagnosis of liver cancer, and the high mortality rate of liver cancer has made it one of the most popular areas for segmentation research. Some deep learning segmentation methods outperformed traditional methods in terms of segmentation results. However, they are unable to obtain satisfactory segmentation results due to blurred original image boundaries, the presence of noise, very small lesion sites, and other factors. In this paper, we propose MDCF_Net, which has dual encoding branches composed of CNN and CnnFormer and can fully utilize multidimensional image features. First, it extracts both intra-slice and inter-slice information and improves the accuracy of the network output by symmetrically using multidimensional fusion layers. In the meantime, we propose a novel feature map stacking approach that focuses on the correlation of adjacent channels of two feature maps, improving the network’s ability to perceive 3D features. Furthermore, the two coding branches collaborate to obtain both texture and edge features, and the network segmentation performance is further improved. Extensive experiments were carried out on the public datasets LiTS to determine the optimal slice thickness for this task. The superiority of the segmentation performance of our proposed MDCF_Net was confirmed by comparison with other leading methods on two public datasets, the LiTS and the 3DIRCADb.

5

Using deep learning to recognize the sign alphabet

Kołodziej Marcin, Szypuła Ernest, Majkowski Andrzej, Rak Remigiusz

Przegląd Elektrotechniczny

|

2022

|

R. 98, nr 6

33--36

EN

This article describes a vision system that uses deep learning to recognize 24 static signs of the American Sign Alphabet in real time. As part of the project, images of signs from four publicly available databases were used as a training set. A DenseNet was implemented for image recognition. For testing, images were acquired with the use of a web camera. The accuracy of sign recognition in images is more than 80%. The real-time version of the system was implemented.

PL

Artykuł zawiera opis systemu wizyjnego wykorzystującego uczenie głębokie do rozpoznawania, w czasie rzeczywistym 24 statycznych znaków Amerykańskiego Alfabetu Migowego. W ramach realizacji projektu, w charakterze zbioru uczącego, wykorzystano obrazy znaków pochodzące z czterech ogólnodostępnych baz danych. Zastosowano sieć DenseNet do rozpoznawania obrazów. Do testów stworzono własne obrazy z wykorzystaniem kamery internetowej. Skuteczność rozpoznawania znaków migowych z wykorzystaniem obrazów przekroczyła 80%. Zaimplementowano wersję systemu pracującą w czasie rzeczywistym.

6

UNet model in image reconstruction for electrical impedance tomography

Maciura Łukasz, Wójcik Dariusz, Rosa Wojciech, Rymarczyk Tomasz, Maj Michał

Przegląd Elektrotechniczny

|

2022

|

R. 98, nr 4

123--126

EN

This paper presents a new algorithm where the UNet convolutional neural network was used to correct deterministic algorithm results, as was is another similar solution using the DBar deterministic algorithm. Instead of the DBar algorithm, another EIT reconstruction algorithm was used in the context cooperation with impedance tomography to extract details in EIT reconstruction. The algorithm uses machine learning to improve the tomographic images obtained with the deterministic algorithm. The final result contains much less noise, and the position of the objects is much better defined, unlike in the deterministic approach. Furthermore, the paper shows how the reconstruction obtained with the hybrid tomograph can be improved to show more details. This paper aims to present a solution that will be used in the context of medical tomography, where the EIT system and the developed algorithm will be used to obtain high-resolution tomography images of the bladder.

PL

Ten artykuł prezentuje nowy algorytm, gdzie sieć konwolucyjna UNet była użyta do korekcji wyników algorytmu deterministycznego jak było w podobnym rozwiązaniu używającym deterministyczny algorytm DBar. Zamiast algorytmu DBar inny algorytm rekonstrukcji EIT został użyty w kontekscie współpracy z tomografią impedancyjną w celu wyodrębnienia szczegółów rekonstrukcji EIT. Algorytm używa uczenie maszynowe do polepszenia obrazów tomograficznych uzyskanych za pomocą algorytmu deterministycznego. Artykuł pokazuje jak rekonstrukcja uzyskana za pomocą tomografu hybrydowego może być ulepszona by ukazywałą więcej szczegółów. Celem tego artykułu jest zaprezentowanie rozwiązania, które będzie użyte w kontekscie tomografii medycznej, gdzie system EIT wraz z którym opracowany algorytm będzie użyty w celu uzyskania wysokiej rozdzielczości obrazów tomograficznych pęcherza moczowego.

7

CCsNeT: Automated Corpus Callosum segmentation using fully convolutional network based on U-Net

Chandra Anjali, Verma Shrish, Raghuvanshi A.S., Bodhey Narendra Kuber

Biocybernetics and Biomedical Engineering

|

2022

|

Vol. 42, no. 1

187--203

EN

Background: Corpus Callosum (CC) is the most prominent white matter bundle in the human brain that connects the left and right cerebral hemispheres. The present paper proposes a novel method for CC segmentation from 2D T1- weighted mid-sagittal brain MRI. The robust segmentation of CC in the mid-sagittal plane plays a vital role in the quantitative study of CC structural features related to various neurological disorders such as Autism, epilepsy, Alzheimer’s disease, and more. Methodology: In this perspective, the current work proposes a Fully Convolutional Network (FCN), a deep learning architecture-based U-Net model for automated CC segmentation from 2D brain MRI images referred to as CCsNeT. The architecture consists of a 35-layers deep, fully convolutional network with two paths, namely contracting and extracting, connected in a U-shape that automatically extracts spatial information. Results: This attempt uses the benchmark brain MRI database comprising ABIDE and OASIS for the experimental investigation. Compared to existing CC segmentation methodologies, the proposed CCsNeT presented improved results achieving Dice Coefficient = 96.74%, and Sensitivity = 97.01% with ABIDE dataset and were further validated against the variants of U-Net model U-Net++, MultiResU-Net, and CE-Net. Further, the performance of CCsNeT has been validated on OASIS and Real-Time Images dataset. Conclusion: Finally, the proposed CCsNeT extracts important CC characteristics such as CC area (CCA) and total brain area (TBA) to categorize the considered 2D MRI slices into control and autism spectrum disorder (ASD) groups, thereby minimizing the inter-observer and intra-observer variability.

8

Deep-segmentation of plantar pressure images incorporating fully convolutional neural networks

Wang Dan, Li Zairan, Dey Nilanjan, Ashour Amira S., Moraru Luminita, Simon Sherratt R., Shi Fuqian

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 1

546--558

EN

Comfort shoe-last design relies on the key points of last curvature. Traditional plantar pressure image segmentation methods are limited to their local and global minimization issues. In this work, an improved fully convolutional networks (FCN) employing SegNet (SegNet+FCN 8 s) is proposed. The algorithm design and operation are performed using the visual geometry group (VGG). The method has high efficiency for the segmentation in positive indices of global accuracy (0.8105), average accuracy (0.8015), and negative indices of average cross-ratio (0.6110) and boundary F1 index (0.6200). The research has potential applications in improving the comfort of shoes.

9

Rozpoznawanie twarzy z wykorzystaniem technik głębokiego uczenia i fuzji danych obrazowych

Jakubowski Jacek, Chmielińska Jolanta

Przegląd Elektrotechniczny

|

2019

|

R. 95, nr 11

150--154

PL

W artykule przedstawiono wyniki oryginalnych badań nad zastosowaniem sieci neuronowej wykorzystującej techniki głębokiego uczenia w zadaniu identyfikacji tożsamości na podstawie obrazów twarzy zarejestrowanych w zakresie widzialnym i w podczerwieni. W badaniach użyte zostały obrazy twarzy eksponowanych w zmiennych ale kontrolowanych warunkach. Na podstawie uzyskanych wyników można stwierdzić, że oba badane zakresy spektralne dostarczają istotnych ale różnych informacji o tożsamości danej osoby, które się wzajemnie uzupełniają.

EN

The paper presents the results of the original research on the application of a neural network using deep learning techniques in the task of identity recognition on the basis of facial images acquired in both visual and thermal radiation ranges. In the research, the database containing images acquired in various but controlled conditions was used. On the basis of the obtained results it can be established that both investigated spectral ranges provide distinctive and complementary details about the identity of an examined person.

10

Koncepcja obiektywizacji testu zegara na potrzeby diagnostyki chorób neurodegeneracyjnych

Jadczak Kamila, Jakubowski Jacek, Tomczykiewicz Kazimierz

Przegląd Elektrotechniczny

|

2019

|

R. 95, nr 11

144--149

PL

W artykule przedstawiona została propozycja wykorzystania inżynierskich technik akwizycji danych i metod głębokiego uczenia do obiektywnej analizy obrazów tworzonych w trakcie wywiadu przez pacjentów z zaburzeniami neurodegeneracyjnymi.

EN

The article presents a proposal of using engineering data acquisition techniques and deep learning methods for an objective analysis of images created during the history-taking in patients with neurodegenerative disorders.

11

Zastosowanie sieci konwolucyjnej głębokiego uczenia w detekcji pojazdów

Oszutowska-Mazurek D. A., Mazurek P.

Problemy Nauk Stosowanych

|

2017

|

T. 7

47-56

PL

Wstęp i cel: Detekcja pojazdów na znaczenie w bezpieczeństwie ruchu drogowego oraz programowaniu pojazdów autonomicznych. Celem pracy jest detekcja pojazdów odróżniająca obrazy pojazdów od innych obrazów nie zawierających pojazdów. Materiał i metody: W pracy wykorzystano bazę pojazdów zawierającą obrazy ekstrahowane z sekwencji wideo, które przetwarzano za pomocą sieci konwolucyjnej głębokiego uczenia. Wyniki: Uzyskana sieć konwolucyjna charakteryzuje się bardzo dobrymi parametrami, krzywa PSNR względem kroku uczenia rośnie co oznacza, że zachodzi proces odszumiania kerneli w całym procesie uczenia. Wniosek: Proponowana metoda może być wykorzystana w programowaniu pojazdów autonomicznych oraz implementacji w Inteligentnych Systemach Transportowych ITS do detekcji pojazdów; bazuje na uczeniu a nie na projektowaniu algorytmu syntetycznego, dzięki temu jest potrzebny relatywnie krótki czas opracowania klasyfikatora.

EN

Introduction and aim: Vehicle detection plays essential role in road safety and automatic vehicle programming. The aim of study is vehicle detection distinguishing car and non-car images Material and methods: Vehicle database images extracted from video sequences were processed by deep learning convolutional network. Results: Obtained convolutional network is characterised by very good parameters, PSNR curve indicates denoising of kernels in learning process. Conclusion: Proposed method is potentially useful in autonomic vehicles programming and Intelligent Transportation Systems (ITS) for vehicles detection. The solution is based on learning, not on synthetic algorithm design, thanks to this, a relatively short time of classifier development is needed.