Search results for: CNN architectures

Authors: Bruno Vilić Belina, Ivan Župan

Abstract:

Smart industry concepts and digital transformation are very popular in many industries. They develop their own digital platforms, which have an important role in innovations and transactions. The main idea of smart industry digital platforms is central data collection, industrial data integration, and data usage for smart applications and services. This paper presents the development of a digital platform for building critical infrastructure protection systems in smart industries. Different service contraction modalities in service level agreements (SLAs), customer relationship management (CRM) relations, trends, and changes in business architectures (especially process business architecture) for the purpose of developing infrastructural production and distribution networks, information infrastructure meta-models and generic processes by critical infrastructure owner demanded by critical infrastructure law, satisfying cybersecurity requirements and taking into account hybrid threats are researched.

Keywords: cybersecurity, critical infrastructure, smart industries, digital platform

Procedia PDF Downloads 74

Authors: Mansouri Nabila, Ben Jemaa Yousra, Motamed Cina, Watelain Eric

Abstract:

Research and development efforts in intelligent Advanced Driver Assistance Systems (ADAS) seek to save lives and reduce the number of on-road fatalities. For traffic and emergency monitoring, the essential but challenging task is vehicle detection and tracking in reasonably short time. This purpose needs first of all a powerful dynamic car detector model. In fact, this paper presents an optimized HOG process based on shape and motion parameters fusion. Our proposed approach mains to compute HOG by bloc feature from foreground blobs using configurable research window and pathway in order to overcome the shortcoming in term of computing time of HOG descriptor and improve their dynamic application performance. Indeed we prove in this paper that HOG by bloc descriptor combined with motion parameters is a very suitable car detector which reaches in record time a satisfactory recognition rate in dynamic outside area and bypasses several popular works without using sophisticated and expensive architectures such as GPU and FPGA.

Keywords: car-detector, HOG, motion, computing time

Procedia PDF Downloads 292

Authors: Christoph Linse, Thomas Martinetz

Abstract:

Recent findings have shown that Neural Networks generalize also in over-parametrized regimes with zero training error. This is surprising, since it is completely against traditional machine learning wisdom. In our empirical study we fortify these findings in the domain of fine-grained image classification. We show that very large Convolutional Neural Networks with millions of weights do learn with only a handful of training samples and without image augmentation, explicit regularization or pretraining. We train the architectures ResNet018, ResNet101 and VGG19 on subsets of the difficult benchmark datasets Caltech101, CUB_200_2011, FGVCAircraft, Flowers102 and StanfordCars with 100 classes and more, perform a comprehensive comparative study and draw implications for the practical application of CNNs. Finally, we show that VGG19 with 140 million weights learns to distinguish airplanes and motorbikes with up to 95% accuracy using only 20 training samples per class.

Keywords: convolutional neural networks, fine-grained image classification, generalization, image recognition, over-parameterized, small data sets

Procedia PDF Downloads 56

Authors: Yiyang Su, Jörg Neumann, Jan Wetzlich, Florian Thiel

Abstract:

Smart grid is a term used to describe the next generation power grid. New challenges such as integration of renewable and decentralized energy sources, the requirement for continuous grid estimation and optimization, as well as the use of two-way flows of energy have been brought to the power gird. In order to achieve efficient, reliable, sustainable, as well as secure delivery of electric power more and more information and communication technologies are used for the monitoring and the control of power grids. Consequently, the need for cybersecurity is dramatically increased and has converged into several standards which will be presented here. These standards for the smart grid must be designed to satisfy both performance and reliability requirements. An in depth investigation of the effect of retrospectively embedded security in existing grids on it’s dynamic behavior is required. Therefore, a retrofitting plan for existing meters is offered, and it’s performance in a test low voltage microgrid is investigated. As a result of this, integration of security measures into measurement architectures of smart grids at the design phase is strongly recommended.

Keywords: cyber security, performance, protocols, security standards, smart grid

Procedia PDF Downloads 295

Authors: Sanjay M. Tailor, Urmila H. Patel

Abstract:

Phthalyl Sulfacetamide belongs to well-known member of antimicrobial sulfonamide family. It is a potent antitumor drug. Structural characteristics of 4-amino-N-(2quinoxalinyl) benzene-sulfonamides (Phthalyl Sulfacetamide), C14H12N4O2S has been studied by method of X-ray crystallography. The compound crystallizes in monoclinic space group P21/n with unit cell parameters a= 7.9841 Ǻ, b= 12.8208 Ǻ, c= 16.6607 Ǻ, α= 90˚, β= 93.23˚, γ= 90˚and Z=4. The X-ray based three-dimensional structure analysis has been carried out by direct methods and refined to an R-value of 0.0419. The crystal structure is stabilized by intermolecular N-H…N, N-H…O and π-π interactions. The Hirshfeld surfaces and consequently the fingerprint analysis have been performed to study the nature of interactions and their quantitative contributions towards the crystal packing. An analysis of Hirshfeld surfaces and fingerprint plots facilitates a comparison of intermolecular interactions, which are the key elements in building different supramolecular architectures. Docking is used for virtual screening for the prediction of the strongest binders based on various scoring functions. Docking studies are carried out on Phthalyl Sulfacetamide for better activity, which is important for the development of a new class of inhibitors.

Keywords: phthalyl sulfacetamide, crystal structure, hirshfeld surface analysis, docking

Procedia PDF Downloads 311

Authors: I. Alhajri, G. Zahedi, R. Alazmi, A. Akbari

Abstract:

In this paper, an Artificial Neural Network (ANN) was developed to predict Asphaltene Precipitation (AP) during the injection of carbon dioxide into crude oil reservoirs. In this study, the experimental data from six different oil fields were collected. Seventy percent of the data was used to develop the ANN model, and different ANN architectures were examined. A network with the Trainlm training algorithm was found to be the best network to estimate the AP. To check the validity of the proposed model, the model was used to predict the AP for the thirty percent of the data that was unevaluated. The Mean Square Error (MSE) of the prediction was 0.0018, which confirms the excellent prediction capability of the proposed model. In the second part of this study, the ANN model predictions were compared with modified Hirschberg model predictions. The ANN was found to provide more accurate estimates compared to the modified Hirschberg model. Finally, the proposed model was employed to examine the effect of different operating parameters during gas injection on the AP. It was found that the AP is mostly sensitive to the reservoir temperature. Furthermore, the carbon dioxide concentration in liquid phase increases the AP.

Keywords: artificial neural network, asphaltene, CO2 injection, Hirschberg model, oil reservoirs

Procedia PDF Downloads 344

Authors: Sean Paulsen, Michael Casey

Abstract:

In this work we present a self-supervised pretraining framework for transformers on functional Magnetic Resonance Imaging (fMRI) data. First, we pretrain our architecture on two self-supervised tasks simultaneously to teach the model a general understanding of the temporal and spatial dynamics of human auditory cortex during music listening. Our pretraining results are the first to suggest a synergistic effect of multitask training on fMRI data. Second, we finetune the pretrained models and train additional fresh models on a supervised fMRI classification task. We observe significantly improved accuracy on held-out runs with the finetuned models, which demonstrates the ability of our pretraining tasks to facilitate transfer learning. This work contributes to the growing body of literature on transformer architectures for pretraining and transfer learning with fMRI data, and serves as a proof of concept for our pretraining tasks and multitask pretraining on fMRI data.

Keywords: transfer learning, fMRI, self-supervised, brain decoding, transformer, multitask training

Procedia PDF Downloads 60

Authors: Babak Forouraghi

Abstract:

A genetic circuit is a collection of interacting genes and proteins that enable individual cells to implement and perform vital biological functions such as cell division, growth, death, and signaling. In cell engineering, synthetic gene circuits are engineered networks of genes specifically designed to implement functionalities that are not evolved by nature. These engineered networks enable scientists to tackle complex problems such as engineering cells to produce therapeutics within the patient's body, altering T cells to target cancer-related antigens for treatment, improving antibody production using engineered cells, tissue engineering, and production of genetically modified plants and livestock. Construction of computational models to realize genetic circuits is an especially challenging task since it requires the discovery of flow of genetic information in complex biological systems. Building synthetic biological models is also a time-consuming process with relatively low prediction accuracy for highly complex genetic circuits. The primary goal of this study was to investigate the utility of a pre-trained bidirectional encoder transformer that can accurately predict gene expressions in genetic circuit designs. The main reason behind using transformers is their innate ability (attention mechanism) to take account of the semantic context present in long DNA chains that are heavily dependent on spatial representation of their constituent genes. Previous approaches to gene circuit design, such as CNN and RNN architectures, are unable to capture semantic dependencies in long contexts as required in most real-world applications of synthetic biology. For instance, RNN models (LSTM, GRU), although able to learn long-term dependencies, greatly suffer from vanishing gradient and low-efficiency problem when they sequentially process past states and compresses contextual information into a bottleneck with long input sequences. In other words, these architectures are not equipped with the necessary attention mechanisms to follow a long chain of genes with thousands of tokens. To address the above-mentioned limitations of previous approaches, a transformer model was built in this work as a variation to the existing DNA Bidirectional Encoder Representations from Transformers (DNABERT) model. It is shown that the proposed transformer is capable of capturing contextual information from long input sequences with attention mechanism. In a previous work on genetic circuit design, the traditional approaches to classification and regression, such as Random Forrest, Support Vector Machine, and Artificial Neural Networks, were able to achieve reasonably high R2 accuracy levels of 0.95 to 0.97. However, the transformer model utilized in this work with its attention-based mechanism, was able to achieve a perfect accuracy level of 100%. Further, it is demonstrated that the efficiency of the transformer-based gene expression classifier is not dependent on presence of large amounts of training examples, which may be difficult to compile in many real-world gene circuit designs.

Keywords: transformers, generative ai, gene expression design, classification

Procedia PDF Downloads 29

Authors: Elif Kalkanli, Constantinos Soutis

Abstract:

The use of textile preform in the advanced fields including aerospace, automotive and marine has exponentially grown in recent years. These preforms offer excellent advantages such as being lightweight and low-cost, and also, their suitability for creating different fiber architectures with different materials whilst improved mechanical properties in certain aspects. In this study, a novel honeycomb core is developed by a 3Dweaving process. The assembly of the layers is achieved thanks to innovative weaving design. Polyester yarn is selected for the 3D woven honeycomb core (3DWHC). The core is used to manufacture a sandwich panel with 2x2 twill glass fiber composite face sheets. These 3DWHC sandwich panels will be tested in three-point bending. The in-plane and out-of-plane (through-the-thickness) mechanical response of the core will be examined as a function of cell size in addition to the flexural response of the sandwich panel. The failure mechanisms of the core and the sandwich skins will be reported in addition to flexural strength and stiffness. Possible engineering applications will be identified.

Keywords: 3D woven, assembly, failure modes, honeycomb sandwich panel

Procedia PDF Downloads 178

Authors: S. Jahani, V. Choopankareh

Abstract:

Many people lost their life caused by environmental pollution every year. The negative effects of environmental crises appear to be much higher in Asian countries. The most important environmental issue in the developing countries and especially in Tehran, to our best knowledge, is air pollution that has affected many aspects of life in society. Environmental topics related to technology’s development have been salient issues among the main concerns of designers. Green facades are the most considerable solutions which designers and architectures are focused on, all over the world. But there are lots of behavioral and psychological problems about this point. In this line, this excavation has tried to reveal the cultural and psychological influences of green façade in developing countries like Tehran. Green façades in developing countries are so useless, although they are so expensive. As a matter of fact, users consider green facade as a decorative item. This research is an attempt to recognize the reasons which show green façades as worthless element. Also, some solutions are presented to promote green façades in the developing countries as an intrinsic solution. There are so many environmental threats, especially about air pollution, for a city as Tehran, which might be solved by green facades.

Keywords: air pollution, developing countries, effects, green facades

Procedia PDF Downloads 242

Authors: Shathya Duobiene, Gediminas Račiukaitis

Abstract:

Solar-powered wireless sensor nodes work best when they operate continuously with minimal energy consumption. Wireless Sensor Networks (WSNs) are a new technology opens up wide studies, and advancements are expanding the prevalence of numerous monitoring applications and real-time aid for environments. The Selective Surface Activation Induced by Laser (SSAIL) technology is an exciting development that gives the design of WSNs more flexibility in terms of their shape, dimensions, and materials. This research work proposes a methodology for using SSAIL technology for forest ecosystem monitoring by wireless sensor networks. WSN monitoring the temperature and humidity were deployed, and their architectures are discussed. The paper presents the experimental outcomes of deploying newly built sensor nodes in forested areas. Finally, a practical method is offered to extend the WSN's lifespan and ensure its continued operation. When operational, the node is independent of the base station's power supply and uses only as much energy as necessary to sense and transmit data.

Keywords: internet of things (IoT), wireless sensor network, sensor nodes, SSAIL technology, forest ecosystem

Procedia PDF Downloads 45

Authors: Mahmood Heshmati, Farhang Daneshmand

Abstract:

Material scientists and engineers have introduced novel materials with complex geometries due to the recent technological advances and promotion of manufacturing methods. Among them, lattice structures with graded architectures denoted by functionally graded porous materials (FGPMs) have been developed to optimize the structural response. FGPMs are achieved by tailoring the size and density of the internal pores in one or more directions that lead to the desired mechanical properties and structural responses. Also, FGPMs provide more flexible transition and the possibility of designing and fabricating structural elements with complex and variable properties. In this paper, wave propagation in lattice structures with functionally graded (FG) porosity is investigated in order to examine the ability of shock absorbing effect. The behavior of FG porous beams with different porosity distributions under impact load and the effects of porosity distribution and porosity content on the wave speed are studied. Important conclusions are made, along with a discussion of the future scope of studies on FGPMs structures.

Keywords: functionally graded, porous materials, wave propagation, impact load, finite element

Procedia PDF Downloads 44

Authors: Babak Forouraghi

Abstract:

A genetic circuit is a collection of interacting genes and proteins that enable individual cells to implement and perform vital biological functions such as cell division, growth, death, and signaling. In cell engineering, synthetic gene circuits are engineered networks of genes specifically designed to implement functionalities that are not evolved by nature. These engineered networks enable scientists to tackle complex problems such as engineering cells to produce therapeutics within the patient's body, altering T cells to target cancer-related antigens for treatment, improving antibody production using engineered cells, tissue engineering, and production of genetically modified plants and livestock. Construction of computational models to realize genetic circuits is an especially challenging task since it requires the discovery of the flow of genetic information in complex biological systems. Building synthetic biological models is also a time-consuming process with relatively low prediction accuracy for highly complex genetic circuits. The primary goal of this study was to investigate the utility of a pre-trained bidirectional encoder transformer that can accurately predict gene expressions in genetic circuit designs. The main reason behind using transformers is their innate ability (attention mechanism) to take account of the semantic context present in long DNA chains that are heavily dependent on the spatial representation of their constituent genes. Previous approaches to gene circuit design, such as CNN and RNN architectures, are unable to capture semantic dependencies in long contexts, as required in most real-world applications of synthetic biology. For instance, RNN models (LSTM, GRU), although able to learn long-term dependencies, greatly suffer from vanishing gradient and low-efficiency problem when they sequentially process past states and compresses contextual information into a bottleneck with long input sequences. In other words, these architectures are not equipped with the necessary attention mechanisms to follow a long chain of genes with thousands of tokens. To address the above-mentioned limitations, a transformer model was built in this work as a variation to the existing DNA Bidirectional Encoder Representations from Transformers (DNABERT) model. It is shown that the proposed transformer is capable of capturing contextual information from long input sequences with an attention mechanism. In previous works on genetic circuit design, the traditional approaches to classification and regression, such as Random Forrest, Support Vector Machine, and Artificial Neural Networks, were able to achieve reasonably high R2 accuracy levels of 0.95 to 0.97. However, the transformer model utilized in this work, with its attention-based mechanism, was able to achieve a perfect accuracy level of 100%. Further, it is demonstrated that the efficiency of the transformer-based gene expression classifier is not dependent on the presence of large amounts of training examples, which may be difficult to compile in many real-world gene circuit designs.

Keywords: machine learning, classification and regression, gene circuit design, bidirectional transformers

Procedia PDF Downloads 33

Authors: Mehdi Assefi, Keihan Hataminezhad

Abstract:

One of the efficient architectures for exploring the resources in wireless ad-hoc networks is contact-based architecture. In this architecture, each node assigns a unique zone for itself and each node keeps all information from inside the zone, as well as some from outside the zone, which is called contact. Reducing the overlap between different zones of a node and its contacts increases its performance, therefore Edge Method (EM) is designed for this purpose. Contacts selected by EM do not have any overlap with their sources, but for choosing the contact a vast amount of information must be transmitted. In this article, we will offer a new protocol for contact selection, which is called PEM. The objective would be reducing the volume of transmitted information, using Non-Uniform Dissemination Probabilistic Protocols. Consumed energy for contact selection is a function of the size of transmitted information between nodes. Therefore, by reducing the content of contact selection message using the PEM will decrease the consumed energy. For evaluation of the PEM we applied the simulation method. Results indicated that PEM consumes less energy compared to EM, and by increasing the number of nodes (level of nodes), performance of PEM will improve in comparison with EM.

Keywords: wireless ad-hoc networks, contact selection, method for CARD, energy-efficient

Procedia PDF Downloads 260

Authors: Mst. Tuhin Akter, Sharun Akter Khushbu, S. M. Shaqib

Abstract:

A real-time system architecture is highly effective for monitoring and detecting various damaged products or fruits that may deteriorate over time or become infected with diseases. Deep learning models have proven to be effective in building such architectures. However, building a deep learning model from scratch is a time-consuming and costly process. A more efficient solution is to utilize deep neural network (DNN) based transfer learning models in the real-time monitoring architecture. This study focuses on using a novel strawberry dataset to develop effective transfer learning models for the proposed real-time monitoring system architecture, specifically for evaluating and detecting strawberry freshness. Several state-of-the-art transfer learning models were employed, and the best performing model was found to be Xception, demonstrating higher performance across evaluation metrics such as accuracy, recall, precision, and F1-score.

Keywords: strawberry freshness evaluation, deep neural network, transfer learning, image augmentation

Procedia PDF Downloads 56

Authors: Ashlesha Khanapure, Harsh Kashyap, Abhinav Anand, Sanjana Habib, Anupama Bidargaddi

Abstract:

Technologies that are developing quickly are being developed daily in a variety of disciplines, particularly the medical field. For the purpose of detecting bone fractures in X-ray pictures of different body segments, our work compares the ResNet-50 and MobileNetV3 architectures. It evaluates accuracy and computing efficiency with X-rays of the elbow, hand, and shoulder from the MURA dataset. Through training and validation, the models are evaluated on normal and fractured images. While ResNet-50 showcases superior accuracy in fracture identification, MobileNetV3 showcases superior speed and resource optimization. Despite ResNet-50’s accuracy, MobileNetV3’s swifter inference makes it a viable choice for real-time clinical applications, emphasizing the importance of balancing computational efficiency and accuracy in medical imaging. We created a graphical user interface (GUI) for MobileNet V3 model bone fracture detection. This research underscores MobileNetV3’s potential to streamline bone fracture diagnoses, potentially revolutionizing orthopedic medical procedures and enhancing patient care.

Keywords: CNN, MobileNet V3, ResNet-50, healthcare, MURA, X-ray, fracture detection

Procedia PDF Downloads 24

Authors: Vinod Kumar Jatav, Dr. Vrijendra Singh

Abstract:

WiMAX is an emerging wireless technology designed by WiMAX forum. WiMAX technology delivers broadband internet access with QoS, mobility and robust security. WiMAX is among the prominent mobile broadband wireless technology which laid the foundation for the next generation networks (NGN). The next-generation communication system for railway should facilitate high level network availability, fast mobility for high speed trains with reliability, high handover rate, the firmness of train operations, and high QoS. The system should also be capable to provide various railway services by transmitting big data efficiently. One of the most promising technologies for the next generation railway wireless communication is Mobile WiMAX. This paper analyses some of the network architectures for railway wireless communication and considers the elementary concepts to facilitate the users with broadband internet access on trains. The paper aims to recognize the suitability of Mobile WiMAX technology for the special requirements of broadband internet facilities and wireless telecommunication services of Railways.

Keywords: Broadband internet, IEEE 802.16e, mobile WiMAX, Railway wireless communication

Procedia PDF Downloads 495

Authors: Güzi̇n Ti̇rkeş, Özge Teki̇n, Kerem Kurtuluş, Y. Yekta Yurtseven, Murat Baran

Abstract:

A scanning electron microscope (SEM) is a type of electron microscope mainly used in nanoscience and nanotechnology areas. Automatic image recognition and classification are among the general areas of application concerning SEM. In line with these usages, the present paper proposes a deep learning algorithm that classifies SEM images into nine categories by means of an online application to simplify the process. The NFFA-EUROPE - 100% SEM data set, containing approximately 21,000 images, was used to train and test the algorithm at 80% and 20%, respectively. Validation was carried out using a separate data set obtained from the Middle East Technical University (METU) in Turkey. To increase the accuracy in the results, the Inception ResNet-V2 model was used in view of the Fine-Tuning approach. By using a confusion matrix, it was observed that the coated-surface category has a negative effect on the accuracy of the results since it contains other categories in the data set, thereby confusing the model when detecting category-specific patterns. For this reason, the coated-surface category was removed from the train data set, hence increasing accuracy by up to 96.5%.

Keywords: convolutional neural networks, deep learning, image classification, scanning electron microscope

Procedia PDF Downloads 84

Authors: Mengqiu Deng, Xiao Peng, Yang Zhao

Abstract:

The development of local energy systems requires the collective involvement of different actors from various levels of society. However, the stakeholder analysis of local energy systems still has been under-developed. This paper proposes an multi-agent system (MAS) framework to facilitate the development of stakeholder analysis of local energy systems. The framework takes into account the most influencing stakeholders, including prosumers/consumers, system operators, energy companies and government bodies. Different stakeholders are modeled based on agent architectures for example the belief-desire-intention (BDI) to better reflect their motivations and interests in participating in local energy systems. The agent models of different stakeholders are then integrated in one model of the whole energy system. An illustrative case study is provided to elaborate how to develop a quantitative agent model for different stakeholders, as well as to demonstrate the practicability of the proposed framework. The findings from the case study indicate that the suggested framework and agent model can serve as analytical instruments for enhancing the government’s policy-making process by offering a systematic view of stakeholder interconnections in local energy systems.

Keywords: multi-agent system, BDI agent, local energy systems, stakeholders

Procedia PDF Downloads 48

Authors: Pablo Chico De Guzman, Cesar Sanchez

Abstract:

Microservice architectures are increasingly in distributed cloud applications due to the advantages on the software composition, development speed, release cycle frequency and the business logic time to market. On the other hand, these architectures also introduce some challenges on the testing and release phases of applications. Container technology solves some of these issues by providing reproducible environments, easy of software distribution and isolation of processes. However, there are other issues that remain unsolved in current container technology when dealing with multiple machines, such as networking for multi-host communication, service discovery, load balancing or data persistency (even though some of these challenges are already solved by traditional cloud vendors in a very mature and widespread manner). Container cluster management tools, such as Kubernetes, Mesos or Docker Swarm, attempt to solve these problems by introducing a new control layer where the unit of deployment is the container (or the pod — a set of strongly related containers that must be deployed on the same machine). These tools are complex to configure and manage and they do not follow a pure immutable infrastructure approach since servers are reused between deployments. Indeed, these tools introduce dependencies at execution time for solving networking or service discovery problems. If an error on the control layer occurs, which would affect running applications, specific expertise is required to perform ad-hoc troubleshooting. As a consequence, it is not surprising that container cluster support is becoming a source of revenue for consulting services. This paper presents i2kit, a deployment tool based on the immutable infrastructure pattern, where the virtual machine is the unit of deployment. The input for i2kit is a declarative definition of a set of microservices, where each microservice is defined as a pod of containers. Microservices are built into machine images using linuxkit —- a tool for creating minimal linux distributions specialized in running containers. These machine images are then deployed to one or more virtual machines, which are exposed through a cloud vendor load balancer. Finally, the load balancer endpoint is set into other microservices using an environment variable, providing service discovery. The toolkit i2kit reuses the best ideas from container technology to solve problems like reproducible environments, process isolation, and software distribution, and at the same time relies on mature, proven cloud vendor technology for networking, load balancing and persistency. The result is a more robust system with no learning curve for troubleshooting running applications. We have implemented an open source prototype that transforms i2kit definitions into AWS cloud formation templates, where each microservice AMI (Amazon Machine Image) is created on the fly using linuxkit. Even though container cluster management tools have more flexibility for resource allocation optimization, we defend that adding a new control layer implies more important disadvantages. Resource allocation is greatly improved by using linuxkit, which introduces a very small footprint (around 35MB). Also, the system is more secure since linuxkit installs the minimum set of dependencies to run containers. The toolkit i2kit is currently under development at the IMDEA Software Institute.

Keywords: container, deployment, immutable infrastructure, microservice

Procedia PDF Downloads 151

Authors: William Whiteley, Jens Gregor

Abstract:

In recent developments of direct neural network based positron emission tomography (PET) reconstruction, two prominent architectures have emerged for converting measurement data into images: 1) networks that contain fully-connected layers; and 2) networks that primarily use a convolutional encoder-decoder architecture. In this paper, we present a multi-modal direct PET reconstruction method called MDPET, which is a hybrid approach that combines the advantages of both types of networks. MDPET processes raw data in the form of sinograms and histo-images in concert with attenuation maps to produce high quality multi-slice PET images (e.g., 8x440x440). MDPET is trained on a large whole-body patient data set and evaluated both quantitatively and qualitatively against target images reconstructed with the standard PET reconstruction benchmark of iterative ordered subsets expectation maximization. The results show that MDPET outperforms the best previously published direct neural network methods in measures of bias, signal-to-noise ratio, mean absolute error, and structural similarity.

Keywords: deep learning, image reconstruction, machine learning, neural network, positron emission tomography

Procedia PDF Downloads 88

Authors: Megha Gupta, Nupur Prakash

Abstract:

Identification of plant diseases has been performed using machine learning and deep learning models on the datasets containing images of healthy and diseased plant leaves. The current study carries out an evaluation of some of the deep learning models based on convolutional neural network (CNN) architectures for identification of plant diseases. For this purpose, the publicly available New Plant Diseases Dataset, an augmented version of PlantVillage dataset, available on Kaggle platform, containing 87,900 images has been used. The dataset contained images of 26 diseases of 14 different plants and images of 12 healthy plants. The CNN models selected for the study presented in this paper are AlexNet, ZFNet, VGGNet (four models), GoogLeNet, and ResNet (three models). The selected models are trained using PyTorch, an open-source machine learning library, on Google Colaboratory. A comparative study has been carried out to analyze the high degree of accuracy achieved using these models. The highest test accuracy and F1-score of 99.59% and 0.996, respectively, were achieved by using GoogLeNet with Mini-batch momentum based gradient descent learning algorithm.

Keywords: comparative analysis, convolutional neural networks, deep learning, plant disease identification

Procedia PDF Downloads 163

Authors: Daniel Gonzales, Zev Winkelman, Trung Tran, Ricardo Sanchez, Dulani Woods, John Hollywood

Abstract:

We have developed a distributed computing capability, Digital Forensics Compute Cluster (DFORC2) to speed up the ingestion and processing of digital evidence that is resident on computer hard drives. DFORC2 parallelizes evidence ingestion and file processing steps. It can be run on a standalone computer cluster or in the Amazon Web Services (AWS) cloud. When running in a virtualized computing environment, its cluster resources can be dynamically scaled up or down using Kubernetes. DFORC2 is an open source project that uses Autopsy, Apache Spark and Kafka, and other open source software packages. It extends the proven open source digital forensics capabilities of Autopsy to compute clusters and cloud architectures, so digital forensics tasks can be accomplished efficiently by a scalable array of cluster compute nodes. In this paper, we describe DFORC2 and compare it with a standalone version of Autopsy when both are used to process evidence from hard drives of different sizes.

Keywords: digital forensics, cloud computing, cyber security, spark, Kubernetes, Kafka

Procedia PDF Downloads 367

Authors: Duy M. P. Vo, Cornelia Sennewald, Gerald Hoffmann, Chokri Cherif

Abstract:

The development of solutions to improve the resistance of buildings to short-term dynamic loads, particularly impact load, is driven by the urgent demand worldwide on securing human life and critical infrastructures. The research training group GRK 2250/1 aims to develop mineral-bonded composites that allow the fabrication of thin-layered strengthening layers providing available concrete members with enhanced impact resistance. This paper presents the development of 3D woven wire cellular structures that can be used as innovative reinforcement for targeted composites. 3D woven wire cellular structures are truss-like architectures that can be fabricated in an automatized process with a great customization possibility. The specific architecture allows this kind of structures to have good load bearing capability and forming behavior, which is of great potential to give strength against impact loading. An appropriate combination of topology and material enables an optimal use of thin-layered reinforcement in concrete constructions.

Keywords: 3D woven cellular structures, ductile behavior, energy absorption, fiber-based reinforced concrete, impact resistant

Procedia PDF Downloads 268

Authors: M. Helen

Abstract:

Chemical graph serves as a convenient model for any real or abstract chemical system. Dendrimers are novel three dimensional hyper branched globular nanopolymeric architectures. Drug delivery scientists are especially enthusiastic about possible utility of dendrimers as drug delivery tool. Dendrimers like poly L lysine (PLL), poly-propylene imine (PPI) and poly-amidoamine (PAMAM), etc., are used as gene carrier in drug delivery system because of their chemical characteristics. These characteristics of chemical compounds are analysed using topological indices (invariants under graph isomorphism) such as Wiener index, Zagreb index, etc., Prof. V. R. Kulli motivated by the application of Zagreb indices in finding the total π energy and derived Gourava indices which is an improved version over Zagreb indices. In this paper, we study the structure of PLL-Dendrimer that has the following applications: reduction in toxicity, colon delivery, and topical delivery. Also, we determine first and second Gourava indices, first and second hyper Gourava indices, product and sum connectivity Gourava indices for PLL-Dendrimer. Gourava Indices have found applications in Quantitative Structure-Property Relationship (QSPR)/ Quantitative Structure-Activity Relationship (QSAR) studies.

Keywords: connectivity Gourava indices, dendrimer, Gourava indices, hyper GouravaG indices

Procedia PDF Downloads 101

Authors: Aminul Haque, Aminul Islam, Prabal Kumar Chowdhury

Abstract:

One of the world’s most pressing issues right now is the lack of a competent waste management system, particularly in emerging and underdeveloped countries. Recycling solid waste, which comprises numerous dangerous non-biodegradable sub-stances like glass, metals, plastics, etc, is the most essential step in reducing waste-related issues in the environment. Typically, collected waste includes all types of waste that must be thoroughly sorted to be recycled efficiently. Most countries use manual waste sorting techniques, which are efficient. Nevertheless, the waste sorting process by human beings is not safe as there is always a risk of exposing themselves to toxic wastes, which could be serious for their health. Our thesis presents a Deep Learning technique based on computer vision for automatically identifying waste. To construct the model, we used Convolutional Neural Networks, real-time object detection systems, such as YOLOv5 and YOLOv7, as well as several transfers learning-based architectures, including VGG16, MobileNet, Inception-Resnet-v2. The model is trained on numerous images for each type of waste to ensure no overfitting and greater accuracy. The highest accuracy we achieved for our waste detection model YOLOv5x, is 93.7%.

Keywords: deep learning, object detection, YOLOv7, image processing, computer vision

Procedia PDF Downloads 19

Authors: Seda Yavuz, Anıl Çelebi, Aysun Taşyapı Çelebi, Oğuzhan Urhan

Abstract:

Nowadays, demand for using real-time video transmission capable devices is ever-increasing. So, high resolution videos have made efficient video compression techniques an essential component for capturing and transmitting video data. Motion estimation has a critical role in encoding raw video. Hence, various motion estimation methods are introduced to efficiently compress the video. Low bit‑depth representation based motion estimation methods facilitate computation of matching criteria and thus, provide small hardware footprint. In this paper, a hardware implementation of a two-bit transformation based low-complexity motion estimation method using local binary pattern approach is proposed. Image frames are represented in two-bit depth instead of full-depth by making use of the local binary pattern as a binarization approach and the binarization part of the hardware architecture is explained in detail. Experimental results demonstrate the difference between the proposed hardware architecture and the architectures of well-known low-complexity motion estimation methods in terms of important aspects such as resource utilization, energy and power consumption.

Keywords: binarization, hardware architecture, local binary pattern, motion estimation, two-bit transform

Procedia PDF Downloads 274

Authors: Muhammad Zubair Khan, Yugyung Lee

Abstract:

Deep learning has recently achieved enormous response in semantic image segmentation. The previously developed U-Net inspired architectures operate with continuous stride and pooling operations, leading to spatial data loss. Also, the methods lack establishing long-term pixels connection to preserve context knowledge and reduce spatial loss in prediction. This article developed encoder-decoder architecture with bi-directional LSTM embedded in long skip-connections and densely connected convolution blocks. The network non-linearly combines the feature maps across encoder-decoder paths for finding dependency and correlation between image pixels. Additionally, the densely connected convolutional blocks are kept in the final encoding layer to reuse features and prevent redundant data sharing. The method applied batch-normalization for reducing internal covariate shift in data distributions. The empirical evidence shows a promising response to our method compared with other semantic segmentation techniques.

Keywords: deep learning, semantic segmentation, image analysis, pixels connection, convolution neural network

Procedia PDF Downloads 76

Authors: Cheikh Bamba Dione, Alla Lo, Elhadji Mamadou Nguer, Siley O. Ba

Abstract:

In this paper, we propose two neural machine translation (NMT) systems (French-to-Wolof and Wolof-to-French) based on sequence-to-sequence with attention and transformer architectures. We trained our models on a parallel French-Wolof corpus of about 83k sentence pairs. Because of the low-resource setting, we experimented with advanced methods for handling data sparsity, including subword segmentation, back translation, and the copied corpus method. We evaluate the models using the BLEU score and find that transformer outperforms the classic seq2seq model in all settings, in addition to being less sensitive to noise. In general, the best scores are achieved when training the models on word-level-based units. For subword-level models, using back translation proves to be slightly beneficial in low-resource (WO) to high-resource (FR) language translation for the transformer (but not for the seq2seq) models. A slight improvement can also be observed when injecting copied monolingual text in the target language. Moreover, combining the copied method data with back translation leads to a substantial improvement of the translation quality.

Keywords: backtranslation, low-resource language, neural machine translation, sequence-to-sequence, transformer, Wolof

Procedia PDF Downloads 114

Authors: Aniket Ganvir, Ritarani Sahu, Suchismita Chinara

Abstract:

The rapid digitization of various aspects of life is leading to the creation of smart IoT ecosystems, where interconnected devices generate significant amounts of valuable data. However, these IoT devices face constraints such as limited computational resources and bandwidth. Cloud computing emerges as a solution by offering ample resources for offloading tasks efficiently despite introducing latency issues, especially for time-sensitive applications like fog computing. Fog computing (FC) addresses latency concerns by bringing computation and storage closer to the network edge, minimizing data travel distance, and enhancing efficiency. Offloading tasks to fog nodes or the cloud can conserve energy and extend IoT device lifespan. The offloading process is intricate, with tasks categorized as full or partial, and its optimization presents an NP-hard problem. Traditional greedy search methods struggle to address the complexity of task offloading efficiently. To overcome this, the efficient crow search algorithm (ECSA) has been proposed as a meta-heuristic optimization algorithm. ECSA aims to effectively optimize computation offloading, providing solutions to this challenging problem.

Keywords: IoT, fog computing, task offloading, efficient crow search algorithm

Procedia PDF Downloads 15