Search results for: convolutional architectures

Authors: Akash Amalan, Rui Wang, Yanqi Qiao, Emmanouil Panaousis, Kaitai Liang

Abstract:

Federated learning is an emerging concept in the domain of distributed machine learning. This concept has enabled General Adversarial Networks (GANs) to benefit from the rich distributed training data while preserving privacy. However, in a non-IID setting, current federated GAN architectures are unstable, struggling to learn the distinct features, and vulnerable to mode collapse. In this paper, we propose an architecture MULTI-FLGAN to solve the problem of low-quality images, mode collapse, and instability for non-IID datasets. Our results show that MULTI-FLGAN is four times as stable and performant (i.e., high inception score) on average over 20 clients compared to baseline FLGAN.

Keywords: federated learning, generative adversarial network, inference attack, non-IID data distribution

Procedia PDF Downloads 112

Authors: Valentina Adduci

Abstract:

The developing research aims to define in general the criteria of environmental sustainability of rural buildings in Puglia and particularly in the manor farm. The main part of the study analyzes the relationship / dependence between the rural building and the landscape which, after many stratifications, results clearly identified and sometimes also characterized in a positive way. The location of the manor farm, in fact, is often conditioned by the infrastructural network and by the structure of the agricultural landscape. The manor farm, without the constraints due to the urban pattern’s density, was developed in accordance with a logical settlement that gives priority to the environmental aspects. These vernacular architectures are the most valuable example of how our ancestors have planned their dwellings according to nature. The 237 farms, analysis’ object, have been reported in cartography through the GIS system; a symbol has been assigned to each of them to identify the architectural typology and a different color for the historical period of construction. A datasheet template has been drawn up, and it has made possible a deeper understanding of each manor farm. This method provides a faster comparison of the most recurring characters in all the considered buildings, except for those farms which benefited from special geographical conditions, such as proximity to the road network or waterways. Below there are some of the most frequently constants derived from the statistical study of the examined buildings: southeast orientation of the main facade; placement of the sheep pen on the ground tilted and exposed to the south side; larger windowed surface on the south elevation; smaller windowed surface on the north elevation; presence of shielding vegetation near the more exposed elevations to the solar radiation; food storage’s rooms located on the ground floor or in the basement; animal shelter located in north side of the farm; presence of tanks and wells, sometimes combined with a very accurate channeling storm water system; thick layers of masonry walls, inside of which were often obtained hollow spaces to house stairwells or depots for the food storage; exclusive use of local building materials. The research aims to trace the ancient use of bioclimatic constructive techniques in the Apulian rural architecture and to define those that derive from an empirical knowledge and those that respond to an already encoded design. These constructive expedients are especially useful to obtain an effective passive cooling, to promote the natural ventilation and to built ingenious systems for the recovery and the preservation of rainwater and are still found in some of the manor farms analyzed, most of them are, today, in a serious state of neglect.

Keywords: bioclimatic devices, farmstead, rural landscape, sustainability

Procedia PDF Downloads 357

Authors: Prachi Desai, Ankita Gandhi, Mitali Acharya

Abstract:

Rainfall forecast is mainly used for predictions of rainfall in a specified area and determining their future rainfall conditions. Rainfall is always a global issue as it affects all major aspects of one's life. Agricultural, fisheries, forestry, tourism industry and other industries are widely affected by these conditions. The studies have resulted in insufficient availability of water resources and an increase in water demand in the near future. We already have a new forecast system that uses the deep Convolutional Neural Network (CNN) to forecast monthly rainfall and climate changes. We have also compared CNN against Artificial Neural Networks (ANN). Machine Learning techniques that are used in rainfall predictions include ARIMA Model, ANN, LR, SVM etc. The dataset on which we are experimenting is gathered online over the year 1901 to 20118. Test results have suggested more realistic improvements than conventional rainfall forecasts.

Keywords: ANN, CNN, supervised learning, machine learning, deep learning

Procedia PDF Downloads 148

Authors: Malgorzata Schwab, Ashis Kumer Biswas

Abstract:

In this concept paper, we explore the topic of Reversibility in Neural Networks leveraged for Network Integrity Verification and crafted the term ''Trusted Neural Network'' (TNN), paired with the API abstraction around it, to embrace the idea formally. This newly proposed high-level generalizable TNN model builds upon the Invertible Neural Network architecture, trained simultaneously in both forward and reverse directions. This allows for the original system inputs to be compared with the ones reconstructed from the outputs in the reversed flow to assess the integrity of the end-to-end inference flow. The outcome of that assessment is captured as an Integrity Score. Concrete implementation reflecting the needs of specific problem domains can be derived from this general approach and is demonstrated in the experiments. The model aspires to become a useful practice in drafting high-level systems architectures which incorporate AI capabilities.

Keywords: trusted, neural, invertible, API

Procedia PDF Downloads 110

Authors: Tsega Asresa, Getahun Tigistu, Melaku Bayih

Abstract:

Computer vision is a subfield of artificial intelligence that allows computers and systems to extract meaning from digital images and video. It is used in a wide range of fields of study, including self-driving cars, video surveillance, medical diagnosis, manufacturing, law, agriculture, quality control, health care, facial recognition, and military applications. Aromatic medicinal plants are botanical raw materials used in cosmetics, medicines, health foods, essential oils, decoration, cleaning, and other natural health products for therapeutic and Aromatic culinary purposes. These plants and their products not only serve as a valuable source of income for farmers and entrepreneurs but also going to export for valuable foreign currency exchange. In Ethiopia, there is a lack of technologies for the classification and identification of Aromatic medicinal plant parts and disease type cured by aromatic medicinal plants. Farmers, industry personnel, academicians, and pharmacists find it difficult to identify plant parts and disease types cured by plants before ingredient extraction in the laboratory. Manual plant identification is a time-consuming, labor-intensive, and lengthy process. To alleviate these challenges, few studies have been conducted in the area to address these issues. One way to overcome these problems is to develop a deep learning model for efficient identification of Aromatic medicinal plant parts with their corresponding disease type. The objective of the proposed study is to identify the aromatic medicinal plant parts and their disease type classification using computer vision technology. Therefore, this research initiated a model for the classification of aromatic medicinal plant parts and their disease type by exploring computer vision technology. Morphological characteristics are still the most important tools for the identification of plants. Leaves are the most widely used parts of plants besides roots, flowers, fruits, and latex. For this study, the researcher used RGB leaf images with a size of 128x128 x3. In this study, the researchers trained five cutting-edge models: convolutional neural network, Inception V3, Residual Neural Network, Mobile Network, and Visual Geometry Group. Those models were chosen after a comprehensive review of the best-performing models. The 80/20 percentage split is used to evaluate the model, and classification metrics are used to compare models. The pre-trained Inception V3 model outperforms well, with training and validation accuracy of 99.8% and 98.7%, respectively.

Keywords: aromatic medicinal plant, computer vision, convolutional neural network, deep learning, plant classification, residual neural network

Procedia PDF Downloads 123

Authors: Supria Sarkar, Vasit Sagan, Sourav Bhadra, Meghnath Pokharel, Felix B.Fritschi

Abstract:

Soybean and their derivatives are very important agricultural commodities around the world because of their wide applicability in human food, animal feed, biofuel, and industries. However, the significance of soybean production depends on the quality of the soybean seeds rather than the yield alone. Seed composition is widely dependent on plant physiological properties, aerobic and anaerobic environmental conditions, nutrient content, and plant phenological characteristics, which can be captured by high temporal resolution remote sensing datasets. Planet scope (PS) satellite images have high potential in sequential information of crop growth due to their frequent revisit throughout the world. In this study, we estimate soybean seed composition while the plants are in the field by utilizing PlanetScope (PS) satellite images and different machine learning algorithms. Several experimental fields were established with varying genotypes and different seed compositions were measured from the samples as ground truth data. The PS images were processed to extract 462 hand-crafted vegetative and textural features. Four machine learning algorithms, i.e., partial least squares (PLSR), random forest (RFR), gradient boosting machine (GBM), support vector machine (SVM), and two recurrent neural network architectures, i.e., long short-term memory (LSTM) and gated recurrent unit (GRU) were used in this study to predict oil, protein, sucrose, ash, starch, and fiber of soybean seed samples. The GRU and LSTM architectures had two separate branches, one for vegetative features and the other for textures features, which were later concatenated together to predict seed composition. The results show that sucrose, ash, protein, and oil yielded comparable prediction results. Machine learning algorithms that best predicted the six seed composition traits differed. GRU worked well for oil (R-Squared: of 0.53) and protein (R-Squared: 0.36), whereas SVR and PLSR showed the best result for sucrose (R-Squared: 0.74) and ash (R-Squared: 0.60), respectively. Although, the RFR and GBM provided comparable performance, the models tended to extremely overfit. Among the features, vegetative features were found as the most important variables compared to texture features. It is suggested to utilize many vegetation indices for machine learning training and select the best ones by using feature selection methods. Overall, the study reveals the feasibility and efficiency of PS images and machine learning for plot-level seed composition estimation. However, special care should be given while designing the plot size in the experiments to avoid mixed pixel issues.

Keywords: agriculture, computer vision, data science, geospatial technology

Procedia PDF Downloads 100

Authors: Anup Anand Deshmukh, Catherine Soladie, Renaud Seguier

Abstract:

Emotion plays a key role in many applications like healthcare, to gather patients’ emotional behavior. Unlike typical ASR (Automated Speech Recognition) problems which focus on 'what was said', it is equally important to understand 'how it was said.' There are certain emotions which are given more importance due to their effectiveness in understanding human feelings. In this paper, we propose an approach that models human stress from audio signals. The research challenge in speech emotion detection is finding the appropriate set of acoustic features corresponding to an emotion. Another difficulty lies in defining the very meaning of emotion and being able to categorize it in a precise manner. Supervised Machine Learning models, including state of the art Deep Learning classification methods, rely on the availability of clean and labelled data. One of the problems in affective computation is the limited amount of annotated data. The existing labelled emotions datasets are highly subjective to the perception of the annotator. We address the first issue of feature selection by exploiting the use of traditional MFCC (Mel-Frequency Cepstral Coefficients) features in Convolutional Neural Network. Our proposed Emo-CNN (Emotion-CNN) architecture treats speech representations in a manner similar to how CNN’s treat images in a vision problem. Our experiments show that Emo-CNN consistently and significantly outperforms the popular existing methods over multiple datasets. It achieves 90.2% categorical accuracy on the Emo-DB dataset. We claim that Emo-CNN is robust to speaker variations and environmental distortions. The proposed approach achieves 85.5% speaker-dependant categorical accuracy for SAVEE (Surrey Audio-Visual Expressed Emotion) dataset, beating the existing CNN based approach by 10.2%. To tackle the second problem of subjectivity in stress labels, we use Lovheim’s cube, which is a 3-dimensional projection of emotions. Monoamine neurotransmitters are a type of chemical messengers in the brain that transmits signals on perceiving emotions. The cube aims at explaining the relationship between these neurotransmitters and the positions of emotions in 3D space. The learnt emotion representations from the Emo-CNN are mapped to the cube using three component PCA (Principal Component Analysis) which is then used to model human stress. This proposed approach not only circumvents the need for labelled stress data but also complies with the psychological theory of emotions given by Lovheim’s cube. We believe that this work is the first step towards creating a connection between Artificial Intelligence and the chemistry of human emotions.

Keywords: deep learning, brain chemistry, emotion perception, Lovheim's cube

Procedia PDF Downloads 123

Authors: Kilari Nikhil, Ankur Tibrewal, Srinivas Kruthiventi S. S.

Abstract:

Conventional speaker identification systems often fall short of capturing the diverse variations present in speech data due to fixed-scale architectures. In this research, we propose a CNN-based architecture, USENet, designed to overcome these limitations. Leveraging two key techniques, our approach achieves superior performance on the VoxCeleb 1 Dataset without any pre-training. Firstly, we adopt a U-net-inspired design to extract features at multiple scales, empowering our model to capture speech characteristics effectively. Secondly, we introduce the squeeze and excitation block to enhance spatial feature learning. The proposed architecture showcases significant advancements in speaker identification, outperforming existing methods, and holds promise for future research in this domain.

Keywords: multi-scale feature extraction, squeeze and excitation, VoxCeleb1 speaker identification, mel-spectrograms, USENet

Procedia PDF Downloads 35

Authors: Vinita Choudhary

Abstract:

Solidly Acoustic Resonators (SMR) based on ZnO piezoelectric material operating at a frequency of 3.96 GHz and 6.49% coupling factor are used to characterize liquids with different viscosities. This behavior of the sensor is analyzed using Finite Element Modeling. Device architectures encapsulate bulk acoustic wave resonators with MO/SiO₂ Bragg mirror reflector and the silicon substrate. The proposed SMR is based on the mass loading effect response of the sensor to the change in the resonant frequency of the resonator that is caused by the increased density due to the absorption of liquids (water, acetone, olive oil) used in theoretical calculation. The sensitivity of sensors ranges from 0.238 MHz/mPa.s to 83.33 MHz/mPa.s, supported by the Kanazawa model. Obtained results are also compared with previous works on BAW viscosity sensors.

Keywords: solidly mounted resonator, bragg mirror, kanazawa model, finite element model

Procedia PDF Downloads 50

Authors: Qiuyu Dai, Haochong Zhang, Xiangrong Liu

Abstract:

Diagonal sparse matrix-vector multiplication is a well-studied topic in the fields of scientific computing and big data processing. However, when diagonal sparse matrices are stored in DIA format, there can be a significant number of padded zero elements and scattered points, which can lead to a degradation in the performance of the current DIA kernel. This can also lead to excessive consumption of computational and memory resources. In order to address these issues, the authors propose the DIA-Adaptive scheme and its kernel, which leverages the parallel instruction sets on MLU. The researchers analyze the effect of allocating a varying number of threads, clusters, and hardware architectures on the performance of SpMV using different formats. The experimental results indicate that the proposed DIA-Adaptive scheme performs well and offers excellent parallelism.

Keywords: adaptive method, DIA, diagonal sparse matrices, MLU, sparse matrix-vector multiplication

Procedia PDF Downloads 77

Authors: Divvela Vishnu Sai Kumar, Deepak Arora, Sheenu Rizvi

Abstract:

Forecasting a cyclone through satellite images consists of the estimation of the intensity of the cyclone and predicting it before a cyclone comes. This research work can help people to take safety measures before the cyclone comes. The prediction of the intensity of a cyclone is very important to save lives and minimize the damage caused by cyclones. These cyclones are very costliest natural disasters that cause a lot of damage globally due to a lot of hazards. Authors have proposed five different CNN (Convolutional Neural Network) models that estimate the intensity of cyclones through INSAT-3D IR images. There are a lot of techniques that are used to estimate the intensity; the best model proposed by authors estimates intensity with a root mean squared error (RMSE) of 10.02 kts.

Keywords: estimating cyclone intensity, deep learning, convolution neural network, prediction models

Procedia PDF Downloads 81

Authors: Sandro Bartolini, Alessandro Mecocci, Alessio Medaglini

Abstract:

The growing popularity of systems based on unmanned aerial vehicles (UAVs) is highlighting their vulnerability, particularly in relation to the positioning system used. Typically, UAV architectures use the civilian GPS, which is exposed to a number of different attacks, such as jamming or spoofing. This is why it is important to develop alternative methodologies to accurately estimate the actual UAV position without relying on GPS measurements only. In this paper, we propose a position estimate method for UAVs based on monocular visual odometry. We have developed a flight control system capable of keeping track of the entire trajectory travelled, with a reduced dependency on the availability of GPS signals. Moreover, the simplicity of the developed solution makes it applicable to a wide range of commercial drones. The final goal is to allow for safer flights in all conditions, even under cyber-attacks trying to deceive the drone.

Keywords: visual odometry, autonomous uav, position measurement, autonomous outdoor flight

Procedia PDF Downloads 186

Authors: S. Vasundra, D. Venkatesh

Abstract:

Wireless networks have gone through an extraordinary growth in the past few years, and will keep on playing a crucial role in future data communication. The present wireless networks aim to make communication possible anywhere and anytime. With the converging of mobile and wireless communications with Internet services, the boundary between mobile personal telecommunications and wireless computer networks is disappearing. Wireless networks of the next generation need the support of all the advances on new architectures, standards, and protocols. Since an ad hoc network may consist of a large number of mobile hosts, this imposes a significant challenge on the design of an effective and efficient routing protocol that can work well in an environment with frequent topological changes. This paper proposes improved ant colony optimization (IACO) technique. It also maintains load balancing in wireless networks. The simulation results show that the proposed IACO performs better than existing routing techniques.

Keywords: wireless networks, ant colony optimization, load balancing, architecture

Procedia PDF Downloads 383

Authors: Aishrith Rao

Abstract:

Human activity recognition is an important aspect of video analytics, and many approaches have been recommended to enable action recognition. In this approach, the model is used to identify the action of the multiple people in the frame and classify them accordingly. A few approaches use RNNs and 3D CNNs, which are computationally expensive and cannot be trained with the small datasets which are currently available. Multi-person action recognition has been performed in order to understand the positions and action of people present in the video frame. The size of the video frame can be adjusted as a hyper-parameter depending on the hardware resources available. OpenPose has been used to calculate pose estimate using CNN to produce heap-maps, one of which provides skeleton features, which are basically joint features. The features are then extracted, and a classification algorithm can be applied to classify the action.

Keywords: human activity recognition, computer vision, pose estimates, convolutional neural networks

Procedia PDF Downloads 108

Authors: Goutam Kumar Ghorai, Sandip Sadhukhan, Arpita Sarkar, Debprasad Sinha, G. Sarkar, Ashis K. Dhara

Abstract:

Diabetes mellitus is one of the most common chronic diseases in all countries and continues to increase in numbers significantly. Diabetic retinopathy (DR) is damage to the retina that occurs with long-term diabetes. DR is a major cause of blindness in the Indian population. Therefore, its early diagnosis is of utmost importance towards preventing progression towards imminent irreversible loss of vision, particularly in the huge population across rural India. The barriers to eye examination of all diabetic patients are socioeconomic factors, lack of referrals, poor access to the healthcare system, lack of knowledge, insufficient number of ophthalmologists, and lack of networking between physicians, diabetologists and ophthalmologists. A few diabetic patients often visit a healthcare facility for their general checkup, but their eye condition remains largely undetected until the patient is symptomatic. This work aims to focus on the design and development of a fully automated intelligent decision system for screening retinal fundus images towards detection of the pathophysiology caused by microaneurysm in the early stage of the diseases. Automated detection of microaneurysm is a challenging problem due to the variation in color and the variation introduced by the field of view, inhomogeneous illumination, and pathological abnormalities. We have developed aconvolutional neural network for efficient detection of microaneurysm. A loss function is also developed to handle severe class imbalance due to very small size of microaneurysms compared to background. The network is able to locate the salient region containing microaneurysms in case of noisy images captured by non-mydriatic cameras. The ground truth of microaneurysms is created by expert ophthalmologists for MESSIDOR database as well as private database, collected from Indian patients. The network is trained from scratch using the fundus images of MESSIDOR database. The proposed method is evaluated on DIARETDB1 and the private database. The method is successful in detection of microaneurysms for dilated and non-dilated types of fundus images acquired from different medical centres. The proposed algorithm could be used for development of AI based affordable and accessible system, to provide service at grass root-level primary healthcare units spread across the country to cater to the need of the rural people unaware of the severe impact of DR.

Keywords: retinal fundus image, deep convolutional neural network, early detection of microaneurysms, screening of diabetic retinopathy

Procedia PDF Downloads 105

Authors: Diego Aguirre, Olac Fuentes

Abstract:

In this paper, we present a taxonomy of weight initialization schemes used in deep learning. We survey the most representative techniques in each class and compare them in terms of overhead cost, convergence rate, and applicability. We also introduce a new weight initialization scheme. In this technique, we perform an initial feedforward pass through the network using an initialization mini-batch. Using statistics obtained from this pass, we initialize the weights of the network, so the following properties are met: 1) weight matrices are orthogonal; 2) ReLU layers produce a predetermined number of non-zero activations; 3) the output produced by each internal layer has a unit variance; 4) weights in the last layer are chosen to minimize the error in the initial mini-batch. We evaluate our method on three popular architectures, and a faster converge rates are achieved on the MNIST, CIFAR-10/100, and ImageNet datasets when compared to state-of-the-art initialization techniques.

Keywords: deep learning, image classification, supervised learning, weight initialization

Procedia PDF Downloads 103

Authors: Samah Benmerbi, Kamal Amroun, Abdelkamel Tari

Abstract:

The increasing number of Web service (WS)providers throughout the globe, have produced numerous Web services providing the same or similar functionality. Therefore, there is a need of tools developing the best answer of queries by selecting and composing services with total transparency. This paper reviews various QoS based Web service selection mechanisms and architectures which facilitate qualitatively optimal selection, in other fact Web service composition is required when a request cannot be fulfilled by a single web service. In such cases, it is preferable to integrate existing web services to satisfy user’s request. We introduce an automatic Web service composition method based on hypergraph decomposition using hypertree decomposition method. The problem of selection and the composition of the web services is transformed into a resolution in a hypertree by exploring the relations of dependency between web services to get composite web service via employing an execution order of WS satisfying global request.

Keywords: web service, web service selection, web service composition, QoS, hypergraph decomposition, BE hypergraph decomposition, hypertree resolution

Procedia PDF Downloads 473

Authors: Aleksei Samarin, Artyom Nazarenko, Valentin Malykh

Abstract:

Image enhancement is becoming more and more popular, especially on mobile devices. Nowadays, it is a common approach to enhance an image using a convolutional neural network (CNN). Such a network should be of significant size; otherwise, a possibility for the artifacts to occur is overgrowing. The existing large CNNs are computationally expensive, which could be crucial for mobile devices. Another important flaw of such models is they are poorly interpretable. There is another approach to image enhancement, namely, the usage of predefined filters in combination with the prediction of their applicability. We present an approach following this paradigm, which outperforms both existing CNN-based and filter-based approaches in the image enhancement task. It is easily adaptable for mobile devices since it has only 47 thousand parameters. It shows the best SSIM 0.919 on RANDOM250 (MIT Adobe FiveK) among small models and is thrice faster than previous models.

Keywords: universal filter, image enhancement, neural networks, computer vision

Procedia PDF Downloads 66

Authors: Abbas Ebrahimi, Mohammad Zandsalimy

Abstract:

The main purpose of this study is to investigate the feasibility of using FPGA (Field Programmable Gate Arrays) chips as alternatives for the conventional CPUs to accelerate the numerical solution of the Laplace equation. FPGA is an integrated circuit that contains an array of logic blocks, and its architecture can be reprogrammed and reconfigured after manufacturing. Complex circuits for various applications can be designed and implemented using FPGA hardware. The reconfigurable hardware used in this paper is an SoC (System on a Chip) FPGA type that integrates both microprocessor and FPGA architectures into a single device. In the present study the Laplace equation is implemented and solved numerically on both reconfigurable hardware and CPU. The precision of results and speedups of the calculations are compared together. The computational process on FPGA, is up to 20 times faster than a conventional CPU, with the same data precision. An analytical solution is used to validate the results.

Keywords: accelerating numerical solutions, CFD, FPGA, hardware definition language, numerical solutions, reconfigurable hardware

Procedia PDF Downloads 353

Authors: Zine-Eddine Bouras, Abdelouaheb Talai

Abstract:

During the last two decades, the software evolution community has intensively tackled the software merging issue whose main objective is to merge in a consistent way different versions of software in order to obtain a new version. Well-established approaches, mainly based on the dependence analysis techniques, have been used to bring suitable solutions. These approaches concern the source code or software architectures. However, these solutions are more expensive due to the complexity and size. In this paper, we overcome this problem by operating at a high level of abstraction. The objective of this paper is to investigate the software merging at the level of UML activity diagrams, which is a new interesting issue. Its purpose is to merge activity diagrams instead of source code. The proposed approach, based on dependence analysis techniques, is illustrated through an appropriate case study.

Keywords: activity diagram, activity diagram slicing, dependency analysis, software merging

Procedia PDF Downloads 295

Authors: Julia Ive, Lucia Specia

Abstract:

Understanding and detecting toxicity is an important problem to support safer human interactions online. Our work focuses on the important problem of contextual toxicity detection, where automated classifiers are tasked with determining whether a short textual segment (usually a sentence) is toxic within its conversational context. We use “toxicity” as an umbrella term to denote a number of variants commonly named in the literature, including hate, abuse, offence, among others. Detecting toxicity in context is a non-trivial problem and has been addressed by very few previous studies. These previous studies have analysed the influence of conversational context in human perception of toxicity in controlled experiments and concluded that humans rarely change their judgements in the presence of context. They have also evaluated contextual detection models based on state-of-the-art Deep Learning and Natural Language Processing (NLP) techniques. Counterintuitively, they reached the general conclusion that computational models tend to suffer performance degradation in the presence of context. We challenge these empirical observations by devising better contextual predictive models that also rely on NLP data augmentation techniques to create larger and better data. In our study, we start by further analysing the human perception of toxicity in conversational data (i.e., tweets), in the absence versus presence of context, in this case, previous tweets in the same conversational thread. We observed that the conclusions of previous work on human perception are mainly due to data issues: The contextual data available does not provide sufficient evidence that context is indeed important (even for humans). The data problem is common in current toxicity datasets: cases labelled as toxic are either obviously toxic (i.e., overt toxicity with swear, racist, etc. words), and thus context does is not needed for a decision, or are ambiguous, vague or unclear even in the presence of context; in addition, the data contains labeling inconsistencies. To address this problem, we propose to automatically generate contextual samples where toxicity is not obvious (i.e., covert cases) without context or where different contexts can lead to different toxicity judgements for the same tweet. We generate toxic and non-toxic utterances conditioned on the context or on target tweets using a range of techniques for controlled text generation(e.g., Generative Adversarial Networks and steering techniques). On the contextual detection models, we posit that their poor performance is due to limitations on both of the data they are trained on (same problems stated above) and the architectures they use, which are not able to leverage context in effective ways. To improve on that, we propose text classification architectures that take the hierarchy of conversational utterances into account. In experiments benchmarking ours against previous models on existing and automatically generated data, we show that both data and architectural choices are very important. Our model achieves substantial performance improvements as compared to the baselines that are non-contextual or contextual but agnostic of the conversation structure.

Keywords: contextual toxicity detection, data augmentation, hierarchical text classification models, natural language processing

Procedia PDF Downloads 134

Authors: X. Gumara Rigol, I. Martínez de Apellaniz Anzuola, A. Garcia Serrano, A. Franzi Cros, O. Vidal Calbet, A. Al Maruf

Abstract:

Traditional federated architectures for data warehousing work well when corporations have existing regional data warehouses and there is a need to aggregate data at a global level. Schibsted Media Group has been maturing from a decentralised organisation into a more globalised one and needed to build both some of the regional data warehouses for some brands at the same time as the global one. In this paper, we present the architectural alternatives studied and why a custom federated approach was the notable recommendation to go further with the implementation. Although the data warehouses are logically federated, the implementation uses a single database system which presented many advantages like: cost reduction and improved data access to global users allowing consumers of the data to have a common data model for detailed analysis across different geographies and a flexible layer for local specific needs in the same place.

Keywords: data integration, data warehousing, federated architecture, Online Analytical Processing (OLAP)

Procedia PDF Downloads 211

Authors: Nouha Khediri, Mohammed Ben Ammar, Monji Kherallah

Abstract:

Recently, facial emotion recognition (FER) has become increasingly essential to understand the state of the human mind. Accurately classifying emotion from the face is a challenging task. In this paper, we present a facial emotion recognition approach named CV-FER, benefiting from deep learning, especially CNN and VGG16. First, the data is pre-processed with data cleaning and data rotation. Then, we augment the data and proceed to our FER model, which contains five convolutions layers and five pooling layers. Finally, a softmax classifier is used in the output layer to recognize emotions. Based on the above contents, this paper reviews the works of facial emotion recognition based on deep learning. Experiments show that our model outperforms the other methods using the same FER2013 database and yields a recognition rate of 92%. We also put forward some suggestions for future work.

Keywords: CNN, deep-learning, facial emotion recognition, machine learning

Procedia PDF Downloads 57

Authors: Rahma Abed, Sahbi Bahroun, Ezzeddine Zagrouba

Abstract:

Due to the huge amount of data in videos, extracting the relevant frames became a necessity and an essential step prior to performing face recognition. In this context, we propose a method for extracting keyframes from videos based on face quality and deep learning for a face recognition task. This method has two steps. We start by generating face quality scores for each face image based on the use of three face feature extractors, including Gabor, LBP, and HOG. The second step consists in training a Deep Convolutional Neural Network in a supervised manner in order to select the frames that have the best face quality. The obtained results show the effectiveness of the proposed method compared to the methods of the state of the art.

Keywords: keyframe extraction, face quality assessment, face in video recognition, convolution neural network

Procedia PDF Downloads 190

Authors: Adeep Hande, Shubham Agarwal

Abstract:

This paper presents a study on identifying sexism in online texts using various state-of-the-art deep learning models based on BERT. We experimented with different feature sets and model architectures and evaluated their performance using precision, recall, F1 score, and accuracy metrics. We also explored the use of pseudolabeling technique to improve model performance. Our experiments show that the best-performing models were based on BERT, and their multilingual model achieved an F1 score of 0.83. Furthermore, the use of pseudolabeling significantly improved the performance of the BERT-based models, with the best results achieved using the pseudolabeling technique. Our findings suggest that BERT-based models with pseudolabeling hold great promise for identifying sexism in online texts with high accuracy.

Keywords: large language models, semi-supervised learning, sexism detection, data sparsity

Procedia PDF Downloads 34

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 269

Authors: Gałka Aleksandra, Jelińska Justyna, Masiak Albert, Walentukiewicz Krzysztof

Abstract:

An offline signature is well-known however not the safest way to verify identity. Nowadays, to ensure proper authentication, i.e. in banking systems, multimodal verification is more widely used. In this paper the online signature analysis based on dynamic time warping (DTW) coupled with machine learning approaches has been presented. In our research signatures made with biometric pens were gathered. Signature features as well as their forgeries have been described. For verification of authenticity various methods were used including convolutional neural networks using DTW matrix and multilayer perceptron using sums of DTW matrix paths. System efficiency has been evaluated on signatures and signature forgeries collected on the same day. Results are presented and discussed in this paper.

Keywords: dynamic time warping, handwritten signature verification, feature-based recognition, online signature

Procedia PDF Downloads 130

Authors: Almagul Assainova , Dariya Abykenova, Liudmila Goncharenko, Sergey Sybachin, Saule Rakhimova, Abay Aman

Abstract:

The recognition of a handwritten Kazakh text is a relevant objective today for the digitization of materials. The study presents a model of a hybrid neural network for handwriting recognition, which includes a convolutional neural network and a multi-layer perceptron. Each network includes 1024 input neurons and 42 output neurons. The model is implemented in the program, written in the Python programming language using the EMNIST database, NumPy, Keras, and Tensorflow modules. The neural network training of such specific letters of the Kazakh alphabet as ә, ғ, қ, ң, ө, ұ, ү, h, і was conducted. The neural network model and the program created on its basis can be used in electronic document management systems to digitize the Kazakh text.

Keywords: handwriting recognition system, image recognition, Kazakh font, machine learning, neural networks

Procedia PDF Downloads 220

Authors: Usama Shahid

Abstract:

Be it written documents, news columns, or students' essays, verifying the content can be a time-consuming task. Apart from the spelling and grammar mistakes, the proofreader is also supposed to verify whether the content included in the essay or document is relevant or not. The irrelevant content in any document or essay is referred to as off-topic text and in this paper, we will address the problem of off-topic text detection from a document using machine learning techniques. Our study aims to identify the off-topic content from a document using Echo state network model and we will also compare data with other models. The previous study uses Convolutional Neural Networks and TFIDF to detect off-topic text. We will rearrange the existing datasets and take new classifiers along with new word embeddings and implement them on existing and new datasets in order to compare the results with the previously existing CNN model.

Keywords: off topic, text detection, eco state network, machine learning

Procedia PDF Downloads 49

Authors: Eman I. Raslan, Haitham S. Hamza, Reda A. El-Khoribi

Abstract:

Fiber-Wireless (FiWi) networks are a promising candidate for future broadband access networks. These networks combine the optical network as the back end where different passive optical network (PON) technologies are realized and the wireless network as the front end where different wireless technologies are adopted, e.g. LTE, WiMAX, Wi-Fi, and Wireless Mesh Networks (WMNs). The convergence of both optical and wireless technologies requires designing architectures with robust efficient and effective bandwidth allocation schemes. Different bandwidth allocation algorithms have been proposed in FiWi networks aiming to enhance the different segments of FiWi networks including wireless and optical subnetworks. In this survey, we focus on the differentiating between the different bandwidth allocation algorithms according to their enhancement segment of FiWi networks. We classify these techniques into wireless, optical and Hybrid bandwidth allocation techniques.

Keywords: fiber-wireless (FiWi), dynamic bandwidth allocation (DBA), passive optical networks (PON), media access control (MAC)

Procedia PDF Downloads 488