Art Convolutional Neural Networks Research Articles

A benchmark for Rey-Osterrieth complex figure test automatic scoring

The Rey–Osterrieth complex figure (ROCF) test is a neuropsychological task that can be useful for early detection of cognitive decline in the elderly population. Several computer vision systems have been proposed to automate this complex analysis task, but the lack of public benchmarks does not allow a fair comparison of these systems. To advance in that direction, we present a benchmarking framework for the automatic scoring of the ROCF test that provides: the ROCFD528 dataset, which is the first open dataset of ROCF line drawings; and experimental results obtained by several modern deep learning models, which can be used as a baseline for comparing new proposals. We evaluate different state-of-the-art convolutional neural networks (CNNs) under traditional and transfer learning paradigms. Experimental quantitative results (MAE = 3.448) indicate that a CNN specifically designed for sketches outperforms other state of the art CNN architectures when the number of examples available is limited. This benchmark can also be a paradigmatic example within the broad field of machine learning for the development of efficient and robust models for analyzing line drawings and sketches not only in classification but also in regression tasks.

НЕЙРОСЕТЕВОЙ ПОДХОД К КОГНИТИВНОМУ МОДЕЛИРОВАНИЮ

Some features of cognitive modeling are presented, including the prerequisites for a cognitive approach to solving complex problems. Cognitive modeling involves the use of various artificial neural networks, including convolutional neural networks. The classification of artificial neural networks according to various characteristics is given. The features of self-organizing neural networks and networks using deep learning methods are considered. The artificial neural network, which is a three-layer unidirectional direct propagation network, the interface of a computer program used to approximate functions using the specified neural network, as well as the solution of the image recognition problem using an artificial convolutional neural network, in which the neural network parameters are adjusted for each recognizable image fragment in order to adaptively filter the image, are considered in detail. The analysis of images in video surveillance systems in order to detect fires allows them to be detected at an early stage and, thus, prevent the fire propogation.

An Effective Approach to SARS-CoV-2 Diagnosis by Developing the CNN Algorithm

Today, SARS is a viral infection of the respiratory system that causes damage to the alveoli. The virus that causes severe acute respiratory syndrome belongs to the family of coronaviruses, commonly known as COVID, which is the same family of viruses that cause common colds. The increasing use of artificial intelligence (AI) and machine learning (ML) nowadays provides excellent opportunities to support and improve medical services in societies. In this paper, a developing model of deep learning (DL) is designed and implemented to classify SARS-CoV-2 disease images into two classes: COVID and non-COVID, with an accuracy of 99%. From a performance and accuracy perspective, the artificial convolutional neural network (CNN)-based detection model that was developed achieved better performance than the pre-trained VGG16 model and other learning models with a similar purpose. As a result, this designed model is able to aid medical professionals by providing a tool that facilitates the detection of this sickness and, consequently, the provision of appropriate medical care.

Wavelength‐Selective Photodetector and Neuromorphic Visual Sensor Utilizing Intrinsic Defect Semiconductor

AbstractWith the rapid developments of Artificial Intelligence (AI) and the Internet of Things (IoT), increasingly intricate and expanding application scenarios are placing higher demands on current machine vision capabilities. Therefore, there is a pressing need to simultaneously achieve diverse functionalities, simple designs, and efficient computing in vision devices. Here, the study develops a two‐terminal optoelectronic device utilizing a single 2D intrinsic defect semiconductor In2S3. The device demonstrates wavelength‐selective photodetection and neuromorphic visual capabilities, attributed to defect‐related charge‐trapping/de‐trapping processes. As a photodetector, the device exhibits a high photoresponsivity of 473.6 A W−1, a high external quantum efficiency of 1.6 × 105%, and a fast rise/fall time of 0.3/1.4 ms at the wavelength of 359 nm. As an all‐in‐one neuromorphic visual device, optoelectronic‐driven fundamental synaptic functions, including paired‐pulse facilitation (PPF), short‐term plasticity (STP), long‐term plasticity (LTP), and “learning‐experience”, are successfully mimicked at the wavelength of 671 nm. Moreover, one‐shot recognition of the 12‐letter image “SHAN XI NORMAL” is achieved through an artificial convolutional neural network. This study provides a new strategy for developing compact high‐level intelligence systems for complex application scenarios.

CONVOLUTIONAL NEURAL NETWORKS USAGE FOR RASTER IMAGE PROCESSING IN NON-DESTRUCTIVE TESTING

An overview of Russian and foreign literature sources on the use of computer vision technologies and convolutional artificial neural networks in technological operations of visual control is presented. Methods and algorithms review of computer vision and artificial neural network models used by the authors of scientific articles, methods of preprocessing and expanding the initial data set for models training under consideration and verifying the correctness of machine vision algorithms is carried out. Possible approaches for automating the assessment of visual quality control in various fields of industry are highlighted, research results, results of machine vision algorithms, neural network models, their accuracy and efficiency indicators are presented, as well as areas and possible scenarios for the application of the presented algorithmic solutions. Some results of machine vision algorithms, neural network models, their accuracy and efficiency indicators are presented.

Design of a flexible tactile sensor for material and texture identification utilizing both contact-separation and surface sliding modes for real-life touch simulation

Haptic feedback plays a crucial role, continuously offering individuals vital information about their external surroundings through physical contact. Therefore, rapid advances in human-friendly biomimetic electronics and flexible devices, have allowed the development of advance electronic skins for artificial haptic sensation. In particular, triboelectric nanogenerator-based tactile sensors have garnered significant interest, pursuing enhanced performance in material and texture recognition, coupled with simplicity in fabrication, cost-effectiveness, high resolution and sensitivity. Thus, in this work, we propose a flexible triboelectric tactile sensor, utilizing the freestanding single-electrode triboelectric nanogenerator principle, realizing concurrent material and texture detection in a single device. We leverage a novel hybrid operational regime for tactile sensing, combining the advantages of both contact-separation and surface sliding modes in a single contact-sliding-separation motion. This enables a realistic representation of the surface employing an artificial finger, by gathering material information from the contact-separation events and texture details from the sliding motion. Common materials such as wood, paper, copper and glass were used for the identification study along with textured surfaces made of the same materials with microscopic surface features. Simultaneous classification and recognition of materials and textures was performed by a single, 1D convolutional artificial neural network, achieving an average recognition accuracy of 98.4% across all different surfaces.

Development of the concept of a multiphase flowmeter

In a number of industries: oil and gas, chemical and nuclear industries, the task of controlling multiphase flow regimes arises. In the nuclear and chemical industries, the flow regime directly affects the nature of technological processes and their safety. In the oil and gas industry, the products extracted from wells are usually a mixture of oil, water and gas, and the task of monitoring the flow regime is related to compliance with the permissible parameters of pumping and control equipment. When using multiphase flowmeters of the flow type, the algorithms for calculating phase flow rates in a multiphase flow are very sensitive to a violation of the uniformity and homogeneity of the measured flow. Excessive noise of the signal of pressure sensors, volume content and flow caused by projectile, cork or stratified modes can negatively affect the accuracy of measurements. As a rule, flow mode maps are used when determining the current mode. This approach is based on the calculation of a number of dimensionless flow parameters (Froude number, Lockhart–Martinelli parameter, etc.). In the case of a dynamically changing flow, this approach may not be suitable. For a more accurate and reliable determination of flow modes, it is proposed to use a direct method of analyzing the spatial distribution of phases in the flow and recognizing the type of flow using artificial convolutional neural networks. This approach allows you to get rid of classification errors and get more accurate information about the flow. The aim of the study is to develop a technique for neural network analysis of images of a multiphase flow with subsequent determination of its type. In the course of the work, approaches to the formation of a training sample are considered, the search for the optimal structure of the neural network is carried out and an accuracy assessment is given for the classification of multiphase flow modes by a convolutional neural network. The study was carried out on two types of data: 1) synthetic images obtained using numerical simulation of multiphase flows, and 2) experimentally obtained flow images on a multiphase flow stand.

Convolutional neural network framework for the automated analysis of transition metal X-ray photoelectron spectra

X-ray photoelectron spectroscopy is an indispensable technique for the quantitative determination of sample composition and electronic structure in diverse research fields. Quantitative analysis of the phases present in XP spectra is usually conducted manually by means of empirical peak fitting performed by trained spectroscopists. However, with recent advancements in the usability and reliability of XPS instruments, ever more (inexperienced) users are creating increasingly large data sets that are harder to analyze by hand. In order to aid users with the analysis of large XPS data sets, more automated, easy-to-use analysis techniques are needed. Here, we propose a supervised machine learning framework based on artificial convolutional neural networks. By training such networks on large numbers of artificially created XP spectra with known quantifications (i.e., for each spectrum, the concentration of each chemical species is known), we created universally applicable models for auto-quantification of transition-metal XPS data that are able to predict the sample composition from spectra within seconds. Upon evaluation against more traditional peak fitting methods, we showed that these neural networks achieve competitive quantification accuracy. The proposed framework is shown to be flexible enough to accommodate spectra containing multiple chemical elements and measured with different experimental parameters. The use of dropout variational inference for the determination of quantification uncertainty is illustrated.

Modern Artificial Neural Networks: Is Evolution Cleverer?

Machine learning tools, particularly artificial neural networks (ANN), have become ubiquitous in many scientific disciplines, and machine learning-based techniques flourish not only because of the expanding computational power and the increasing availability of labeled data sets but also because of the increasingly powerful training algorithms and refined topologies of ANN. Some refined topologies were initially motivated by neuronal network architectures found in the brain, such as convolutional ANN. Later topologies of neuronal networks departed from the biological substrate and began to be developed independently as the biological processing units are not well understood or are not transferable to in silico architectures. In the field of neuroscience, the advent of multichannel recordings has enabled recording the activity of many neurons simultaneously and characterizing complex network activity in biological neural networks (BNN). The unique opportunity to compare large neuronal network topologies, processing, and learning strategies with those that have been developed in state-of-the-art ANN has become a reality. The aim of this review is to introduce certain basic concepts of modern ANN, corresponding training algorithms, and biological counterparts. The selection of these modern ANN is prone to be biased (e.g., spiking neural networks are excluded) but may be sufficient for a concise overview.

Development of Software Module for Recognizing Traffic Flows Through Deep Learning

This work describes a software system that uses machine learning methods and classifies vehicles observed in a frame. Analysis of the subject area and the possibilities provided by the methods of machine learning in the framework of the performance of assigned tasks are realized. An important part of the work is devoted to a study of the features of machine learning technologies with a teacher. In particular, the basic principles of the functioning of convolutional artificial neural networks are considered. This paper describes the implementation of the final software module, and its functionality, and demonstrates the results of its testing. Results from the developed software module show significant improvement in the accuracy over other methods, such as Artificial Neural Networks. Further developments of this technology may lead to broader applications and further improvements.

Machine learning of cloud types in satellite observations and climate models

Abstract. Uncertainty in cloud feedbacks in climate models is a major limitation in projections of future climate. Therefore, evaluation and improvement of cloud simulation are essential to ensure the accuracy of climate models. We analyse cloud biases and cloud change with respect to global mean near-surface temperature (GMST) in climate models relative to satellite observations and relate them to equilibrium climate sensitivity, transient climate response and cloud feedback. For this purpose, we develop a supervised deep convolutional artificial neural network for determination of cloud types from low-resolution (2.5∘×2.5∘) daily mean top-of-atmosphere shortwave and longwave radiation fields, corresponding to the World Meteorological Organization (WMO) cloud genera recorded by human observers in the Global Telecommunication System (GTS). We train this network on top-of-atmosphere radiation retrieved by the Clouds and the Earth’s Radiant Energy System (CERES) and GTS and apply it to the Coupled Model Intercomparison Project Phase 5 and 6 (CMIP5 and CMIP6) model output and the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis version 5 (ERA5) and the Modern-Era Retrospective Analysis for Research and Applications version 2 (MERRA-2) reanalyses. We compare the cloud types between models and satellite observations. We link biases to climate sensitivity and identify a negative linear relationship between the root mean square error of cloud type occurrence derived from the neural network and model equilibrium climate sensitivity (ECS), transient climate response (TCR) and cloud feedback. This statistical relationship in the model ensemble favours models with higher ECS, TCR and cloud feedback. However, this relationship could be due to the relatively small size of the ensemble used or decoupling between present-day biases and future projected cloud change. Using the abrupt-4×CO2 CMIP5 and CMIP6 experiments, we show that models simulating decreasing stratiform and increasing cumuliform clouds tend to have higher ECS than models simulating increasing stratiform and decreasing cumuliform clouds, and this could also partially explain the association between the model cloud type occurrence error and model ECS.

State of the Art Convolutional Neural Networks

State of the Art Convolutional Neural Networks

НЕЙРОСЕТЕВОЙ АНАЛИЗ ЦИФРОВЫХ ИЗОБРАЖЕНИЙ ЛЕЙКОЦИТОВ ДЛЯ ОБНАРУЖЕНИЯ ЗАБОЛЕВАНИЙ КРОВИ

The article discusses the application of the transfer learning method for the ensemble of artificial convolutional neural networks with preliminary digital image segmentation for blood cells in order to classify them later. The results obtained during neural networks classification demonstrate the efficiency of such technologies used for improving the accuracy of artificial neural networks when solving the problems of medical images segmentation for leukocytes in order to diagnose hematologic diseases.

Memory-Throughput Trade-off for CNN-Based Applications at the Edge

Many modern applications require execution of Convolutional Neural Networks (CNNs) on edge devices, such as mobile phones or embedded platforms. This can be challenging, as the state-of-the art CNNs are memory costly, whereas the memory budget of edge devices is highly limited. To address this challenge, a variety of CNN memory reduction methodologies have been proposed. Typically, the memory of a CNN is reduced using methodologies such as pruning and quantization. These methodologies reduce the number or precision of CNN parameters, thereby reducing the CNN memory cost. When more aggressive CNN memory reduction is required, the pruning and quantization methodologies can be combined with CNN memory reuse methodologies. The latter methodologies reuse device memory allocated for storage of CNN intermediate computational results, thereby further reducing the CNN memory cost. However, the existing memory reuse methodologies are unfit for CNN-based applications that exploit pipeline parallelism available within the CNNs or use multiple CNNs to perform their functionality. In this article, we therefore propose a novel CNN memory reuse methodology. In our methodology, we significantly extend and combine two existing CNN memory reuse methodologies to offer efficient memory reuse for a wide range of CNN-based applications.

Robust and Responsive Learning of Spatiotemporal Urban Traffic Flow Relationships

Spatiotemporal models that incorporate information about the relationships between traffic flows over space and time have recently become the mainstream of urban traffic forecasting. Many algorithms have been developed to learn the structure of spatiotemporal relationships – the travel time-based, cross-correlation, graphical lasso, mutual information, and random forest methods, among several others. In this paper, we address two issues of spatiotemporal relationship learning. First, we suggest the application of ensemble techniques (ranking- and voting-based) that combine the advantages of individual algorithms and allow more robust structures to be obtained for learning spatiotemporal relationships. Second, we propose strategies that enable discovery of their dynamics and provide extensive information on traffic flows for the responsive learning of spatiotemporal relationships. In addition to the widely used forecasting performance (estimated for the spatially regularised vector autoregression model, the spatiotemporal k-nearest neighbour algorithm, and the high-order graph convolutional artificial neural network), we compare the estimated spatiotemporal relationship structures using their complexity and stability metrics and demonstrate the importance of these characteristics. A multitude of numerical experiments using a large citywide traffic data set are conducted, and the results demonstrate the preferences of ensemble learning and the strengths of the proposed learning strategies.

DIAGNOSIS OF BRAIN TUMOR USING MULTISCALE CONVOLUTION NEURAL NETWORK

Nowadays, the number of patients with brain tumors is steadily increasing, diagnosis and isolation of the tumor play an important role in the process of treatment and surgery. Due to the high error of manual segmentation of the tumor, algorithms that perform this operation with less error are of great importance. Convolutional neural networks have made great progress in the field of medical imaging. The use of imaging techniques and pattern recognition in the diagnosis and automatic determination of brain tumors by MRI imaging reduces errors, human error and speeds up detection. The artificial convolutional neural network (CNN) has been widely used in the diagnosis of intelligent cancers and has significantly reduced the error rate. Therefore, in this paper, we present a new method using a combination of convolutional and multi-scale artificial neural network that has significantly increased the accuracy of tumor diagnosis. This study presents a multidisciplinary convolution neural network (MCNN) approach to classifying tumors that can be used as an important part of automated diagnosis systems for accurate cancer diagnosis. Based on the MCNN structure, which presents the MRI image to several deep convolutional neural networks of varying sizes and resolutions, the stage of extracting classical hand-made features is avoided. This approach proposes better classification rates than the classical methods. This study uses a multi-scale convolution technique to achieve a detection accuracy of 95/4%, which shows the efficiency of the proposed method.

Interpretable Artificial Intelligence through Locality Guided Neural Networks

In current deep learning architectures, each of the deeper layers in networks tends to contain hundreds of unorganized neurons which makes it hard for humans to understand how they interact with each other. By organizing the neurons using correlation as the criteria, humans can observe how clusters of neighbouring neurons interact with each other. Research in Explainable Artificial Intelligence (XAI) aims to all alleviate the black-box nature of current AI methods and make them understandable by humans. In this paper, we extend our previous algorithm for XAI in deep learning, called Locality Guided Neural Network (LGNN). LGNN preserves locality between neighbouring neurons within each layer of a deep network during training. Motivated by Self-Organizing Maps (SOMs), the goal is to enforce a local topology on each layer of a deep network such that neighbouring neurons are highly correlated with each other. Our algorithm can be easily plugged into current state of the art Convolutional Neural Network (CNN) models to make the neighbouring neurons more correlated. A cluster of neighbouring neurons activating for a class makes the network both quantitatively and qualitatively more interpretable when visualized, as we show through our experiments. This paper focuses on image processing with CNNs, but can theoretically be applied to any type of deep learning architecture. In our experiments, we train VGG and WRN networks for image classification on CIFAR100 and Imagenette. Our experiments analyse different perceptible clusters of activations in response to different input classes.

Machine learning – An approach for consistent rock glacier mapping and inventorying – Example of Austria

Rock glaciers (RG) are landforms that occur in high latitudes or elevations and — in their active state — consist of a mixture of rock debris and ice. Despite serving as a form of groundwater storage, they are an indicator for the occurrence of (former) permafrost and therefore carry significance in the research for the ongoing climate change. For these reasons, the past years have shown rising interest in the establishment of RG inventories to investigate the extent of permafrost and quantify water storages. Creating these inventories, however, usually involves manual, laborious, and subjective mapping of the landforms based on aerial image - and digital elevation model analysis. We propose an approach for RG mapping based on supervised machine learning which can help to increase the mapping efficiency and permits rapid RG mapping in vast and not yet covered areas. We found deep convolutional artificial neural networks (ANN) that are specifically designed for image segmentation (U-Net architecture) to be well suited for this classification problem. The general workflow consists of training the ANNs with orthophotos and slope maps of digital elevation models as input. The output (RG label-maps) is derived from a recently published RG inventory of the Austrian Alps that features 5769 individual RGs and was compiled manually by several scientists. To increase the generalization capabilities, we use live data augmentation during training. Based on this inventory, the ANNs have learned the average expert opinion and the RG map generated by the ANN can be used to increase the consistency and completeness of already existing RG inventories. Moreover, this ANN approach might be valuable for other landform mapping tasks beyond rock glaciers (e.g., other mass movements).

Machine and Deep Learning in the Evaluation of Selected Qualitative Characteristics of Sweet Potatoes Obtained under Different Convective Drying Conditions

This paper discusses the use of various methods to distinguish between slices of sweet potato dried in different conditions. The drying conditions varied in terms of temperature, the values were: 60 °C, 70 °C, 80 °C, and 90 °C. Examination methods included instrumental texture analysis using a texturometer and digital texture analysis based on macroscopic images. Classification of acquired data involved the use of machine learning techniques using various types of artificial neural networks, such as convolutional neural networks (CNNs) and multi-layer perceptron (MLP). As a result, in the convective drying, changes in color darkening were found in products with the following temperature values: 60 °C (L = 83.41), 70 °C (L = 81.11), 80 °C (L = 79.02), and 90 °C (L = 75.53). The best-generated model achieved an overall classification efficiency of 77%. Sweet potato dried at 90 °C proved to be completely distinguishable from other classes, among which classification efficiency varied between 61–83% depending on the class. This means that image analysis using deep convolutional artificial neural networks is a valuable tool in the context of assessing the quality of convective-dried sweet potato slices.

Deep Learning Based a New Approach for Power Quality Disturbances Classification in Power Transmission System

Power quality is one of the most important research eras for the energy sector. Suddenly dropped voltages or suddenly rising voltages and harmonics in energy should be identified. All of these distortions are called power quality disturbances (PQDs). Deep learning based convolutional artificial neural networks with an attention model approach has been carried out. The main idea is to develop a new approach to convolutional neural network (CNN) based which classifies a particular power signal into its respective power quality condition. The attention model approach is based on the idea that the best solution will be taken from the newly produced data pool obtained by rescaling the available data according to the total number of pixels before the average data pool is created and then deep CNN processes will continue. In the attention model approach, all data is multiplied by the number of elements by the number of epoch time sixty-six tensors. The dataset used here contains signals which belong to one of the 9 classes. This means that each signal is characterized by 622 data points and 5600 data parameters. All signals provided are in time domain. Power quality (PQ) is directly depending on power disturbances’ absence or scarcity. The accuracy and error values of the developed model were obtained according to both the number of epochs and the number of iterations.