Search results for: fused deep representations

Authors: Boumaraf Malak

Abstract:

The development of intelligent transportation systems (ITS) and artificial intelligence (AI) are spurring us to pave the way for the widespread adoption of autonomous vehicles (AVs). This is open again opportunities for smart roads, smart traffic safety, and mobility comfort. A highly intelligent decision-making system is essential for autonomous driving around dense, dynamic objects. It must be able to handle complex road geometry and topology, as well as complex multiagent interactions, and closely follow higher-level commands such as routing information. Autonomous vehicles have become a very hot research topic in recent years due to their significant ability to reduce traffic accidents and personal injuries. Using new artificial intelligence-based technologies handles important functions in scene understanding, motion planning, decision making, vehicle control, social behavior, and communication for AV. This paper focuses only on deep reinforcement learning-based methods; it does not include traditional (flat) planar techniques, which have been the subject of extensive research in the past because reinforcement learning (RL) has become a powerful learning framework now capable of learning complex policies in high dimensional environments. The DRL algorithm used so far found solutions to the four main problems of autonomous driving; in our paper, we highlight the challenges and point to possible future research directions.

Keywords: deep reinforcement learning, autonomous driving, deep deterministic policy gradient, deep Q-learning

Procedia PDF Downloads 85

Authors: Ramezani M., Neitzert T.

Abstract:

This paper investigates mechanical properties and formability of the AZ61 magnesium alloy at high temperatures. Tensile tests were performed at elevated temperatures of up to 400ºC. The results showed that as temperature increases, yield strength and ultimate tensile strength decrease significantly, while the material experiences an increase in ductility (maximum elongation before break). A finite element model has been developed to further investigate the formability of the AZ61 alloy by deep drawing a square cup. Effects of different process parameters such as punch and die geometry, forming speed and temperature as well as blank-holder force on deep drawability of the AZ61 alloy were studied and optimum values for these parameters are achieved which can be used as a design guide for deep drawing of this alloy.

Keywords: AZ61, formability, magnesium, mechanical properties

Procedia PDF Downloads 579

Authors: Eunji Choi, Borang Park, Youngjae Choi, Jinwoo Moon

Abstract:

In this study, a predictive model for estimating the metabolism (MET) of human body was developed for the optimal control of indoor thermal environment. Human body images for indoor activities and human body joint coordinated values were collected as data sets, which are used in predictive model. A deep learning algorithm was used in an initial model, and its number of hidden layers and hidden neurons were optimized. Lastly, the model prediction performance was analyzed after the model being trained through collected data. In conclusion, the possibility of MET prediction was confirmed, and the direction of the future study was proposed as developing various data and the predictive model.

Keywords: deep learning, indoor quality, metabolism, predictive model

Procedia PDF Downloads 257

Authors: Niklas Panten, Eberhard Abele

Abstract:

This paper presents a novel approach for real-time and near-optimal control of industrial smart grids by deep reinforcement learning (DRL). To achieve highly energy-efficient factory systems, the energetic linkage of machines, technical building equipment and the building itself is desirable. However, the increased complexity of the interacting sub-systems, multiple time-variant target values and stochastic influences by the production environment, weather and energy markets make it difficult to efficiently control the energy production, storage and consumption in the hybrid industrial smart grids. The studied deep reinforcement learning approach allows to explore the solution space for proper control policies which minimize a cost function. The deep neural network of the DRL agent is based on a multilayer perceptron (MLP), Long Short-Term Memory (LSTM) and convolutional layers. The agent is trained within multiple Modelica-based factory simulation environments by the Advantage Actor Critic algorithm (A2C). The DRL controller is evaluated by means of the simulation and then compared to a conventional, rule-based approach. Finally, the results indicate that the DRL approach is able to improve the control performance and significantly reduce energy respectively operating costs of industrial smart grids.

Keywords: industrial smart grids, energy efficiency, deep reinforcement learning, optimal control

Procedia PDF Downloads 195

Authors: Jun Ming Moey, Zhiyaun Chen, David Nicholson

Abstract:

Bad actors are often hard to detect in data that imprints their behaviour patterns because they are comparatively rare events embedded in non-bad actor data. An unsupervised machine learning framework is applied here to detect bad actors in financial crime datasets that record millions of transactions undertaken by hundreds of actors (<0.01% bad). Specifically, the framework combines ‘shallow’ (PCA, Isolation Forest) and ‘deep’ (Autoencoder) methods to detect outlier patterns. Detection performance analysis for both the individual methods and their combination is reported.

Keywords: detection, machine learning, deep learning, unsupervised, outlier analysis, data science, fraud, financial crime

Procedia PDF Downloads 94

Authors: Nasser Ghassembaglou, Abdullah Bolek, Oktay Yilmaz, Ertan Oznergiz, Hikmet Kocabas, Safak Yilmaz

Abstract:

High quality air filters production using nanofibers, as a functional material, has frequently been investigated. As it is more environmentally friendly, melting method has been selected to produce nanofibers. Spun-bond production systems consist of extruder, spin-pump, nozzle package and attenuators. Spin-pump makes molten polymer steady, which flows through extruder. Fibers are formed by regular melts passing through nuzzle holes under high pressure. Attenuator prolongs fibers to micron size to be collected on a conveyor. Different designs of attenuator systems have been studied in this research; new analysis have been done on existed designs considering fibers effect on air flow; it was comprehended that, at fibers presence, there is an air flow which agglomerates fibers as a negative effect. So some new representations have been designed and CFD analysis have been done on them. Afterwards, one of these representations selected as the most optimum and effective design which is brought in this paper.

Keywords: attenuator, CFD, nanofiber, spun-bond

Procedia PDF Downloads 448

Authors: Haozhe Xiang

Abstract:

With the rise of the Internet of Things era, intelligent products are gradually integrating into people's lives. Pedestrian trajectory prediction has become a key issue, which is crucial for the motion path planning of intelligent agents such as autonomous vehicles, robots, and drones. In the current technological context, deep learning technology is becoming increasingly sophisticated and gradually replacing traditional models. The pedestrian trajectory prediction algorithm combining neural networks and attention mechanisms has significantly improved prediction accuracy. Based on in-depth research on deep learning and pedestrian trajectory prediction algorithms, this article focuses on physical environment modeling and learning of historical trajectory time dependence. At the same time, social interaction between pedestrians and scene interaction between pedestrians and the environment were handled. An improved pedestrian trajectory prediction algorithm is proposed by analyzing the existing model architecture. With the help of these improvements, acceptable predicted trajectories were successfully obtained. Experiments on public datasets have demonstrated the algorithm's effectiveness and achieved acceptable results.

Keywords: deep learning, graph convolutional network, attention mechanism, LSTM

Procedia PDF Downloads 70

Authors: Gaokai Liu

Abstract:

Automatic defect detection for nanomaterial images is widely required in industrial scenarios. Deep learning approaches are considered as the most effective solutions for the great majority of image-based tasks. In this paper, an edge guidance network for defect segmentation is proposed. First, the encoder path with multiple convolution and downsampling operations is applied to the acquisition of shared features. Then two decoder paths both are connected to the last convolution layer of the encoder and supervised by the edge and segmentation labels, respectively, to guide the whole training process. Meanwhile, the edge and encoder outputs from the same stage are concatenated to the segmentation corresponding part to further tune the segmentation result. Finally, the effectiveness of the proposed method is verified via the experiments on open nanofibrous datasets.

Keywords: deep learning, defect detection, image segmentation, nanomaterials

Procedia PDF Downloads 149

Authors: Li Qingjian, Li Ke, He Chun, Huang Yong

Abstract:

In this paper, the method of combining the Pohl Seidman's deep belief network with the self-organizing neural network is proposed to classify the target. This method is mainly aimed at the high nonlinearity of the hyperspectral image, the high sample dimension and the difficulty in designing the classifier. The main feature of original data is extracted by deep belief network. In the process of extracting features, adding known labels samples to fine tune the network, enriching the main characteristics. Then, the extracted feature vectors are classified into the self-organizing neural network. This method can effectively reduce the dimensions of data in the spectrum dimension in the preservation of large amounts of raw data information, to solve the traditional clustering and the long training time when labeled samples less deep learning algorithm for training problems, improve the classification accuracy and robustness. Through the data simulation, the results show that the proposed network structure can get a higher classification precision in the case of a small number of known label samples.

Keywords: DBN, SOM, pattern classification, hyperspectral, data compression

Procedia PDF Downloads 341

Authors: Mikyung Bak

Abstract:

The dokkaebi, a beloved icon of Korean culture, is represented as an elf, goblin, monster, dwarf, or any similar creature in different media, such as animated shows, comics, soap operas, and movies. It is often described as a mythical creature with a horn or horns and long teeth, wearing tiger-skin pants or a grass skirt, and carrying a magic stick. Many Korean researchers agree on the similarity of the image of the Korean dokkaebi with that of the Japanese oni, a view that is regard as negative from an anti-colonial or nationalistic standpoint. They cite such similarity between the two mythical creatures as evidence that Japanese colonialism persists in Korea. The debate on the originality of dokkaebi’s visual representation is an issue that must be addressed urgently. This research demonstrates through a diagram the plurality of interpretations of dokkaebi’s visual representations in what are considered ‘authentic’ images of dokkaebi in Korean art and culture. This diagram presents the opinions of four major groups in the debate, namely, the scholars of Korean literature and folklore, art historians, authors, and artists. It also shows the creation of new dokkaebi visual representations in popular media, including those influenced by the debate. The diagram further proves that dokkaebi’s representations varied, which include the typical persons or invisible characters found in Korean literature, original Korean folk characters in traditional art, and even universal spirit characters. They are also visually represented by completely new creatures as well as oni-based mythical beings and the actual oni itself. The earlier dokkaebi representations were driven by the creation of a national ideology or national cultural paradigm and, thus, were more uniform and protected. In contrast, the more recent representations are influenced by the Korean industrial strategy of ‘cultural economics,’ which is concerned with the international rather than the domestic market. This recent Korean cultural strategy emphasizes diversity and commonality with the global culture rather than originality and locality. It employs traditional cultural resources to construct a global image. Consequently, dokkaebi’s recent representations have become more common and diverse, thereby incorporating even oni’s characteristics. This argument has rendered the grounds of the debate irrelevant. The dokkaebi has been used recently for tourist marketing purposes, particularly in revitalizing interest in regions considered the cradle of various traditional dokkaebi tales. These campaign strategies include the Jeju-do Dokkaebi Park, Koksung Dokkaebi Land, as well as the Taebaek and Sokri-san Dokkaebi Festivals. Almost dokkaebi characters are identical to the Japanese oni in tourist marketing. However, the pursuit for dokkaebi’s authentic visual representation is less interesting and fruitful than the appreciation of the entire spectrum of dokkaebi images that have been created. Thus, scholars and stakeholders must not exclude the possibilities for a variety of potentials within the visual culture. The same sentiment applies to traditional art and craft. This study aims to contribute to a new visualization of the dokkaebi that embraces the possibilities of both folk craft and art, which continue to be uncovered by diverse and careful researchers in a still-developing field.

Keywords: Dokkaebi, post-colonial period, representation, tourist marketing

Procedia PDF Downloads 278

Authors: Agoubi Belgacem, Adel Kharroubi

Abstract:

Geothermal groundwater is the main water source to supply various sectors in El Hamma city, southeastern Tunisia. This region was long the destination of thousands of people from Tunisia and neighboring countries for care and bathing. The main objective of this study is to understand the groundwater mineralization origins and factors that control. The second goal is the appraisal of radon in geothermal groundwater in the study area. For this aim, geothermal groundwater was sampled and collected from different locations (thermal baths and deep wells). Physical parameters were measured and major ions were analyzed. Results reveal three water types. The water first type has Na-Mg-Ca-SO4-Cl facies and T>55°C. The second water type dominated by Na-Ca-Cl-SO4 facies with a temperature < 45 °C. However the third water type is dominated by Ca-SO4-Na-Cl-Mg. The three water types may be controlled by depth and geology. The first represent groundwater from deep aquifer (lower cretaceous), the second type was the shallow aquifer and the first is mixed water from deep and shallow water with a temperature ranging from 45 to 55°C. Measured Radon shows that shallow aquifer has a higher 222Rn concentration (677 to 2903 Bq.m-3) than deep water (203 to 1100 Bq.m-3). R-222 in El Hamma thermal aquifer was controlled by structures, porosity and permeability of aquifers. Geostatistical analyses of hydrogeological data and radon activities confirm the vertical flow and communication between deep and shallow aquifers through vertical faults system.

Keywords: Radon-222, geothermal, water, environment, Tunisia

Procedia PDF Downloads 360

Authors: Jim J. Catina, Jackee M. Gwynn, Jin S. Kang

Abstract:

Additive manufacturing (AM) for hybrid rocket engines is becoming increasingly attractive due to its ability to create complex grain configurations with improved regression rates when compared to cast grains. However, the presence of microvoids in parts produced through the additive manufacturing method of Fused Deposition Modeling (FDM) results in a lower fuel density and is believed to cause a decrease in regression rate compared to ideal performance. In this experiment, FDM was used to create hybrid rocket fuel grains with a star configuration composed of acrylonitrile butadiene styrene (ABS). Testing was completed to determine the effect of heat treatment as a post-processing method to improve the combustion performance of hybrid rocket fuel grains manufactured by FDM. For control, three ABS star configuration grains were printed using FDM and hot fired using gaseous oxygen (GOX) as the oxidizer. Parameters such as thrust and mass flow rate were measured. Three identical grains were then heat treated to varying degrees and hot fired under the same conditions as the control grains. This paper will quantitatively describe the amount of improvement in engine performance as a result of heat treatment of the AM hybrid fuel grain. Engine performance is measured in this paper by specific impulse, which is determined from the thrust measurements collected in testing.

Keywords: acrylonitrile butadiene styrene, additive manufacturing, fused deposition modeling, heat treatment

Procedia PDF Downloads 117

Authors: Jawad Abuhlail, Hamza Hroub

Abstract:

Let R be a commutative ring. Representation theory studies the representation of R-modules as (possibly finite) sums of special types of R-submodules. Here we are interested in a class of R-modules between the class of semisimple R-modules and the class of R-modules that can be written as (possibly finite) sums of secondary R-submodules (we know that every simple R-submodule is secondary). We investigate R-modules which can be written as (possibly finite) sums of second R-submodules (we call those modules second representable). Moreover, we investigate the class of (main) second attached prime ideals related to a module with such representation. We provide sufficient conditions for an R-module M to get a (minimal) second representation. We also found the collection of second attached prime ideals for some types of second representable R-modules, in particular within the class of injective R-modules. As we know that every simple R-submodule is second and every second R-submodule is secondary, we can see the importance of the second representable R-module.

Keywords: lifting modules, second attached prime ideals, second representations, secondary representations, semisimple modules, second submodules

Procedia PDF Downloads 192

Authors: Shunzuo Wu, Xudong Luo, Yuanxiu Liao

Abstract:

Recently, the rapid development of deep learning makes artificial intelligence (AI) penetrate into many fields, replacing manual work there. In particular, AI systems also become a research focus in the field of automatic office. To meet real needs in automatic officiating, in this paper we develop an automatic form filling system. Specifically, it uses two classical neural network models and several word embedding models to classify various relevant information elicited from the Internet. When training the neural network models, we use less noisy and balanced data for training. We conduct a series of experiments to test my systems and the results show that our system can achieve better classification results.

Keywords: artificial intelligence and office, NLP, deep learning, text classification

Procedia PDF Downloads 200

Authors: Brinda Sharma, Saakshi S. Sarpotdar

Abstract:

Deep-fat frying is popular standard method that has been studied basically to clarify the complicated mechanisms of fat decomposition at high temperatures and to assess their effects on human health. The aim of this paper is to point out the application of method engineering that has been recently improved our understanding of the fundamental principles and mechanisms concerned at different scales and different times throughout the process: pretreatment, frying, and cooling. It covers the several aspects of deep-fat drying. New results regarding the understanding of the frying method that are obtained as a results of major breakthroughs in on-line instrumentation (heat, steam flux, and native pressure sensors), within the methodology of microstructural and imaging analysis (NMR, MRI, SEM) and in software system tools for the simulation of coupled transfer and transport phenomena. Such advances have opened the approach for the creation of significant information of the behavior of varied materials and to the event of latest tools to manage frying operations via final product quality in real conditions. Lastly, this paper promotes an integrated approach to the frying method as well as numerous competencies like those of chemists, engineers, toxicologists, nutritionists, and materials scientists also as of the occupation and industrial sectors.

Keywords: frying, cooling, imaging analysis (NMR, MRI, SEM), deep-fat frying

Procedia PDF Downloads 430

Authors: T. Zaki, Fathi S. Soliman

Abstract:

Desulfurization process is required by most, if not all refineries, to achieve ultra-low sulfur fuel, that contains less than 10 ppm sulfur. A lot of research works and many effective technologies have been studied to achieve deep desulfurization process in moderate reaction environment, such as adsorption desulfurization (ADS), oxidative desulfurization (ODS), biodesulfurization and extraction desulfurization (EDS). Extraction desulfurization using deep eutectic solvents (DESs) is considered as simple, cheap, highly efficient and environmentally friend process. In this work, four DESs were designed and synthesized. Choline chloride (ChCl) was selected as typical hydrogen bond acceptors (HBA), and ethylene glycol (EG), glycerol (Gl), urea (Ur) and thiourea (Tu) were selected as hydrogen bond donors (HBD), from which a series of deep eutectic solvents were synthesized. The experimental data showed that the synthesized DESs showed desulfurization affinities towards the thiophene species in cyclohexane solvent. Ethylene glycol molecules showed more affinity to create hydrogen bond with thiophene instead of choline chloride. Accordingly, ethylene glycol choline chloride DES has the highest extraction efficiency.

Keywords: DES, desulfurization, green solvent, extraction

Procedia PDF Downloads 288

Authors: Gizem Eser Erdek

Abstract:

This study is research on predicting the remaining life of industrial cutting tools used in the industrial production process with deep learning methods. When the life of cutting tools decreases, they cause destruction to the raw material they are processing. This study it is aimed to predict the remaining life of the cutting tool based on the damage caused by the cutting tools to the raw material. For this, hole photos were collected from the hole-drilling machine for 8 months. Photos were labeled in 5 classes according to hole quality. In this way, the problem was transformed into a classification problem. Using the prepared data set, a model was created with convolutional neural networks, which is a deep learning method. In addition, VGGNet and ResNet architectures, which have been successful in the literature, have been tested on the data set. A hybrid model using convolutional neural networks and support vector machines is also used for comparison. When all models are compared, it has been determined that the model in which convolutional neural networks are used gives successful results of a %74 accuracy rate. In the preliminary studies, the data set was arranged to include only the best and worst classes, and the study gave ~93% accuracy when the binary classification model was applied. The results of this study showed that the remaining life of the cutting tools could be predicted by deep learning methods based on the damage to the raw material. Experiments have proven that deep learning methods can be used as an alternative for cutting tool life estimation.

Keywords: classification, convolutional neural network, deep learning, remaining life of industrial cutting tools, ResNet, support vector machine, VggNet

Procedia PDF Downloads 76

Authors: Dingyang Hu, Dan Liu

Abstract:

DNN (Deep Neural Network) deep learning models are widely used in classification, prediction, and other task scenarios. To address the difficulties of generic adversarial perturbation generation for deep learning models under black-box conditions, a generic adversarial ingestion generation method based on a saliency map (CJsp) is proposed to obtain salient image regions by counting the factors that influence the input features of an image on the output results. This method can be understood as a saliency map attack algorithm to obtain false classification results by reducing the weights of salient feature points. Experiments also demonstrate that this method can obtain a high success rate of migration attacks and is a batch adversarial sample generation method.

Keywords: adversarial sample, gradient, probability, black box

Procedia PDF Downloads 104

Authors: Yasemin İlkay

Abstract:

Street is not only a crucial spatial unit in urban design and planning discipline but also the context of walking practice in urban space. Moreover, psychogeography concentrates on both ‘walking’ and, therefore, the differentiated forms of (urban) streets to examine the influence of the built environment on the feelings and attitudes of human beings. This paper focuses on ‘walking practice’ in university campuses with reference to spatial appropriation forms via a psychogeographic lens on the phenomenon of alle in two different cities of Turkey, Ankara, the capital city, and Van, in the eastern part of the country. Alle, as an extension of ‘street’ in university campuses, is the constructive spatial structure in university campuses, and as a result, it should be the (both physical and mental) spine of design policy while conceiving and constructing a university campus. The main question of the paper is: How does the interrelation of ‘campus design’ and ‘walking practice’ on alle penetrate reciprocally on the spatial representations of citizens within their urban daily lives. The body contacts with and at urban space (with other objects and subjects) via its movements and stops; this interaction occurs through the spatial pattern of occupancy and vacancy. Walking practice leads to a set of cognitive mental representations in relation to the repertoire of place attachment and spatial appropriation. University campuses are autonomous and fruitful urban spaces to investigate such an interaction. There are both physical/real and psychogeographic representations of the same urban spaces and urban spatial practices. This separation would indicate the invisible dimensions of the difference between ‘what is conceived’ and ‘what is perceived.’ This study aims to compare and contrast the role of alle in both campus design and spatial appropriation via walking at two differentiated university campuses by collecting the mental representations, doing in-depth interviews, and attending walks with the interviewees by psychogeographic techniques. Campus design and spatial appropriation will be compared [with reference to the conception and perception of alle] in three scales: (1) the historical spatial development stories and design approaches of university campuses, (2) the spatial pattern of campuses on the basis of alle, and (3) sub-behavioral regions of the alle in campuses in relation with mental representations and psychogeographic attentive walks. The sub-questions of the research are: [1] How and why do the design approaches differentiate in two university campuses in Turkey, [2] How the interrelation among alle design and spatial appropriation differs in these two cases, and [3] What do the differentiated gaps among real and psychographic maps indicate about the design and spatial appropriation interrelation. METU, as a well-designed, readable campus with its alle, promise a rich walking practice with in-depth and fruitful spatial appropriation regions; however, Van YYÜ limits both the practice and place attachment with its partial design with an alle which is later added to the campus. This research both displays the role of alle in the campus design, walking practice and spatial appropriation and opens a new methodological path to discover hidden knowledge within urban spaces.

Keywords: alle, campus design, cognitive geography, psychogeography, spatial appropriation, Turkey

Procedia PDF Downloads 101

Authors: Iván Rivet, Narges Dialami, Miguel Cervera, Michele Chiumenti

Abstract:

Process parameters in Additive Manufacturing (AM) play a critical role in the mechanical performance of the final component. In order to find the input configuration that guarantees the optimal performance of the printed part, the process-performance relationship must be found. Fused Filament Fabrication (FFF) is the selected demonstrative AM technology due to its great popularity in the industrial manufacturing world. A material model that considers the different printing patterns present in a FFF part is used. A voxelized mesh is built from the manufacturing toolpaths described in the G-Code file. An Adaptive Mesh Refinement (AMR) based on the octree strategy is used in order to reduce the complexity of the mesh while maintaining its accuracy. High-fidelity and cost-efficient Finite Element (FE) simulations are performed and the influence of key process parameters in the mechanical performance of the component is analyzed. A robust optimization process based on appropriate failure criteria is developed to find the printing direction that leads to the optimal mechanical performance of the component. The Tsai-Wu failure criterion is implemented due to the orthotropy and heterogeneity constitutive nature of FFF components and because of the differences between the strengths in tension and compression. The optimization loop implements a modified version of an Anomaly Detection (AD) algorithm and uses the computed metrics to obtain the optimal printing direction. The developed methodology is verified with a case study on an industrial demonstrator.

Keywords: additive manufacturing, optimization, printing direction, mechanical performance, voxelization

Procedia PDF Downloads 63

Authors: Raz Flieshman, Adam Michael Altenbuchner, Jörg Krüger

Abstract:

This paper presents an approach to optimizing the Fused Deposition Modeling (FDM) 3D printing process through a preprocess calibration routine of printing parameters. The core of this method involves the use of a low-cost thermal sensor capable of measuring tempera-tures within the range of -20 to 500 degrees Celsius for detailed process observation. The calibration process is conducted by printing a predetermined path while varying the process parameters through machine instructions (g-code). This enables the extraction of critical thermal, dimensional, and surface properties along the printed path. The calibration routine utilizes computer vision models to extract features and metrics from the thermal images, in-cluding temperature distribution, layer adhesion quality, surface roughness, and dimension-al accuracy and consistency. These extracted properties are then analyzed to optimize the process parameters to achieve the desired qualities of the printed material. A significant benefit of this calibration method is its potential to create printing parameter profiles for new polymer and composite materials, thereby enhancing the versatility and application range of FDM 3D printing. The proposed method demonstrates significant potential in enhancing the precision and reliability of FDM 3D printing, making it a valuable contribution to the field of additive manufacturing.

Keywords: FDM 3D printing, preprocess calibration, thermal sensor, process optimization, additive manufacturing, computer vision, material profiles

Procedia PDF Downloads 40

Authors: Ons Slimene, Aroua Taamallah, Maha Khemaja

Abstract:

In recent years, human activity recognition (HAR) has been a key area of research due to its diverse applications. It has garnered increasing attention in the field of computer vision. HAR plays an important role in people’s daily lives as it has the ability to learn advanced knowledge about human activities from data. In HAR, activities are usually represented by exploiting different types of sensors, such as embedded sensors or visual sensors. However, these sensors have limitations, such as local obstacles, image-related obstacles, sensor unreliability, and consumer concerns. Recently, several deep learning-based approaches have been proposed for HAR and these approaches are classified into two categories based on the type of data used: vision-based approaches and sensor-based approaches. This research paper highlights the importance of multimodal data fusion from skeleton data obtained from videos and data generated by embedded sensors using deep neural networks for achieving HAR. We propose a deep multimodal fusion network based on a twostream architecture. These two streams use the Convolutional Neural Network combined with the Bidirectional LSTM (CNN BILSTM) to process skeleton data and data generated by embedded sensors and the fusion at the feature level is considered. The proposed model was evaluated on a public OPPORTUNITY++ dataset and produced a accuracy of 96.77%.

Keywords: human activity recognition, action recognition, sensors, vision, human-centric sensing, deep learning, context-awareness

Procedia PDF Downloads 101

Authors: Yuanqing Li, Changyi Lei, Zhaopeng Xue, Zhuo Zheng, Yanbo Long

Abstract:

In agricultural production and breeding, implementing automatic picking robot in orchard farming to reduce human labour and error is challenging. The core function of it is automatic identification based on machine vision. This paper focuses on apple detection and counting in orchards and implements several deep learning methods. Extensive datasets are used and a semi-automatic annotation method is proposed. The proposed deep learning models are in state-of-the-art YOLO family. In view of the essence of the models with various backbones, a multi-dimensional comparison in details is made in terms of counting accuracy, mAP and model memory, laying the foundation for realising automatic precision agriculture.

Keywords: agricultural object detection, deep learning, machine vision, YOLO family

Procedia PDF Downloads 197

Authors: Tayeb Aissaoui, Inas M. AlNashef

Abstract:

In recent years, green solvents named deep eutectic solvents (DESs) have been found to possess significant properties and to be applicable in several technologies. Choline chloride (ChCl) mixed with urea at a ratio of 1:2 and 80 °C was the first discovered DES. In this article, chemical structure and combination mechanism of ChCl: urea based DES were investigated. Moreover, the implementation of this DES in water removal from natural gas was reported. Dehydration of natural gas by ChCl:urea shows significant absorption efficiency compared to triethylene glycol. All above operations were retrieved from COSMOthermX software. This article confirms the potential application of DESs in gas industry.

Keywords: COSMO-RS, deep eutectic solvents, dehydration, natural gas, structure, organic salt

Procedia PDF Downloads 292

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 198

Authors: Anastasios Ntousas

Abstract:

The purpose of the following paper is to analyze the relationship between the lexical quality and the self-teaching hypothesis and their impact on the reading ability. The following questions emerged, is there a correlation between the effective reading experience that the lexical quality hypothesis proposes and the self-teaching hypothesis, would the ability to read by analogy facilitate and create stable, synchronized four-word representational, and would word morphological knowledge be a possible extension of the self-teaching hypothesis. The lexical quality hypothesis speculates that words include four representational attributes, phonology, orthography, morpho-syntax, and meaning. Those four-word representations work together to make word reading an effective task. A possible lack of knowledge in one of the representations might disrupt reading comprehension. The degree that the four-word features connect together makes high and low lexical word quality representations. When the four-word representational attributes connect together effectively, readers have a high lexical quality of words; however, when they hardly have a strong connection with each other, readers have a low lexical quality of words. Furthermore, the self-teaching hypothesis proposes that phonological recoding enables printed word learning. Phonological knowledge and reading experience facilitate the acquisition and consolidation of specific-word orthographies. The reading experience is related to strong reading comprehension. The more readers have contact with texts, the better readers they become. Therefore, their phonological knowledge, as the self-teaching hypothesis suggests, might have a facilitative impact on the consolidation of the orthographical, morphological-syntax and meaning representations of unknown words. The phonology of known words might activate effectively the rest of the representational features of words. Readers use their existing phonological knowledge of similarly spelt words to pronounce unknown words; a possible transference of this ability to read by analogy will appear with readers’ morphological knowledge. Morphemes might facilitate readers’ ability to pronounce and spell new unknown words in which they do not have lexical access. Readers will encounter unknown words with similarly phonemes and morphemes but with different meanings. Knowledge of phonology and morphology might support and increase reading comprehension. There was a careful selection, discussion of theoretical material and comparison of the two existing theories. Evidence shows that morphological knowledge improves reading ability and comprehension, so morphological knowledge might be a possible extension of the self-teaching hypothesis, the fundamental skill to read by analogy can be implemented to the consolidation of word – specific orthographies via readers’ morphological knowledge, and there is a positive correlation between effective reading experience and self-teaching hypothesis.

Keywords: morphology, orthography, reading ability, reading comprehension

Procedia PDF Downloads 128

Authors: T. Damak, O. Kriaa, A. Baccar, M. A. Ben Ayed, N. Masmoudi

Abstract:

In the last few years, Automatic Number Plate Recognition (ANPR) systems have become widely used in the safety, the security, and the commercial aspects. Forethought, several methods and techniques are computing to achieve the better levels in terms of accuracy and real time execution. This paper proposed a computer vision algorithm of Number Plate Localization (NPL) and Characters Segmentation (CS). In addition, it proposed an improved method in Optical Character Recognition (OCR) based on Deep Learning (DL) techniques. In order to identify the number of detected plate after NPL and CS steps, the Convolutional Neural Network (CNN) algorithm is proposed. A DL model is developed using four convolution layers, two layers of Maxpooling, and six layers of fully connected. The model was trained by number image database on the Jetson TX2 NVIDIA target. The accuracy result has achieved 95.84%.

Keywords: ANPR, CS, CNN, deep learning, NPL

Procedia PDF Downloads 306

Authors: Shuo Wang

Abstract:

The hydrogeological parameters of the engineering site and the hydraulic connection between the aquifers can be obtained by the pumping test. Through the recovery test, the characteristics of water level recovery and the law of surface subsidence recovery can be understood. The above two tests can provide the basis for subsequent engineering design. At present, the deformation of deep soil caused by pumping tests is often neglected. However, some studies have shown that the maximum settlement subject to groundwater drawdown is not necessarily on the surface but in the deep soil. In addition, the law of settlement recovery of each soil layer subject to water level recovery is not clear. If the deformation-sensitive structure is deep in the test site, safety accidents may occur. In this study, the pumping test and recovery test of a confined aquifer in Shanghai are introduced. The law of measured groundwater changes and surface subsidence are analyzed. In addition, the fluid-solid coupling model was established by ABAQUS based on the Biot consolidation theory. The models are verified by comparing the computed and measured results. Further, the variation law of water level and the deformation law of deep soil during pumping and recovery tests under different site conditions and different times and spaces are discussed through the above model. It is found that the maximum soil settlement caused by pumping in a confined aquifer is related to the permeability of the overlying aquitard and pumping time. There is a lag between soil deformation and groundwater changes, and the recovery rate of settlement deformation of each soil layer caused by the rise of water level is different. Finally, some possible research directions are proposed to provide new ideas for academic research in this field.

Keywords: coupled hydro-mechanical analysis, deep ground settlement, numerical simulation, pumping test, recovery test

Procedia PDF Downloads 44

Authors: Smitha Haridasan, Zelalem Demissie, Atri Dutta, Ajita Rattani

Abstract:

Aided by the wind, all it takes is one ember and a few minutes to create a wildfire. Wildfires are growing in frequency and size due to climate change. Wildfires and its consequences are one of the major environmental concerns. Every year, millions of hectares of forests are destroyed over the world, causing mass destruction and human casualties. Thus early detection of wildfire becomes a critical component to mitigate this threat. Many computer vision-based techniques have been proposed for the early detection of forest fire using video surveillance. Several computer vision-based methods have been proposed to predict and detect forest fires at various spectrums, namely, RGB, HSV, and YCbCr. The aim of this paper is to propose a multi-spectral deep learning model that combines information from different spectrums at intermediate layers for accurate fire detection. A heterogeneous dataset assembled from publicly available datasets is used for model training and evaluation in this study. The experimental results show that multi-spectral deep learning models could obtain an improvement of about 4.68 % over those based on a single spectrum for fire detection.

Keywords: deep learning, forest fire detection, multi-spectral learning, natural hazard detection

Procedia PDF Downloads 241

Authors: Ijaz Ahmed, Anwar Zada, Muhammad Waqas, M. U. Ashraf

Abstract:

We demonstrate the performance of a very efficient tagger applies on hadronically decaying top quark pairs as signal based on deep neural network algorithms and compares with the QCD multi-jet background events. A significant enhancement of performance in boosted top quark events is observed with our limited computing resources. We also compare modern machine learning approaches and perform a multivariate analysis of boosted top-pair as well as single top quark production through weak interaction at √s = 14 TeV proton-proton Collider. The most relevant known background processes are incorporated. Through the techniques of Boosted Decision Tree (BDT), likelihood and Multlayer Perceptron (MLP) the analysis is trained to observe the performance in comparison with the conventional cut based and count approach

Keywords: top tagger, multivariate, deep learning, LHC, single top

Procedia PDF Downloads 111