Pētniecība
Publikācijas
729 publikācijas
- 2022
CNN for Hand Washing Movement Classification: What Matters More-the Approach or the Dataset?(atveras jaunā logā)
Atis Elsts, Maksims Ivanovs, Roberts Kadikis, Olegs Sabelnikovs
2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)
Kopsavilkums
Good hand hygiene is one of the key factors in preventing infectious diseases, including COVID-19. Advances in machine learning have enabled automated hand hygiene evaluation, with research papers reporting highly accurate hand washing movement classification from video data. However, existing studies typically use datasets collected in lab conditions. In this paper, we apply state-of-the-art techniques such as MobileNetV2 based CNN, including two-stream and recurrent CNN, to three different datasets: a good-quality and uniform lab-based dataset, a more diverse lab-based dataset, and a large-scale real-life dataset collected in a hospital. The results show that while many of the approaches show good accuracy on the first dataset, the accuracy drops significantly o n t he m ore complex datasets. Moreover, all approaches fail to generalize on the third dataset, and only show slightly-better-than random accuracy on videos held out from the training set. This suggests that despite the high accuracy routinely reported in the research literature, the transition to real-world applications for hand washing quality monitoring is not going to be straightforward.
- 2022
Applicability of different neural network architectures in UWB signal processing for different object classification(atveras jaunā logā)
Kristaps Greitans, Modris Greitans
2022 Workshop on Microwave Theory and Techniques in Wireless Communications (MTTW), lpp. 138-143, IEEE
Kopsavilkums
In the paper, the performance of different Artificial neural network (ANN) architectures - CNN, RNN, Transformers, CNN/LSTM, ResNet, and MLP is discussed in UWB impulse radio radar signal classification. The signals are obtained by reflecting and passing UWB pulses through different material objects that give different information for the classification purpose. The ANN architectures are compared on their classification precision, training time, and memory requirements. The training data consists of 144 objects including regular and crashed PET bottles, glass bottles, and metal cans. The results show the accuracy's of classification if mono-static (reflected signals are analyzed), bistatic (propagated trough signals), and multi-static setups of UWB radar are used. For single-channel cases GRU (99.65%), Resnet (99.69%), Transformer (99.66%) architectures are preferable, while the highest multi-static evaluation accuracy reaches 99.90% for the Transformer. As expected the more dense ANN networks perform better classification.
- 2022
Influence of different types of real-time feedback on hand washing quality assessed with neural networks/simulated neural networks(atveras jaunā logā)
O. Zemlanuhina, M. Lulla, A. Rutkovskis, A. Slavinska, A. Vilde, A. Melbarde-Kelmere, A. Elsts, M. Ivanovs, O. Sabelnikovs
SHS Web of Conferences, Sēj. 131, lpp. 02008, EDP Sciences
Kopsavilkums
Background: Thousands of people die every day around the world from infections acquired in a hospital. Hands are the main pathways of germ transmission during healthcare. Hand hygiene monitoring can be performed using various methods. One of the latest techniques that can combine all is a neural network-based hand hygiene monitoring system. Methods/Design: Each participant performed 3 hand-washing trials, each time receiving different type of feedback. The order in which each participant of the study used the developed applications was strictly defined, thus each hand-washing study session started with performing hand washing using application A, B and C accordingly. All captured videos of hand-wash episodes were saved and later analysed with neural networks. In the end, both evaluation results were compared and evaluated. Results show that when the participants use Application Type A, they perform hand washing much faster, as well as in comparison of Application Type A versus application type C. However, the longest time spent for the hand washing was detected while using the application type B. Conclusion: Study shows that structured guidance provided during the real time hand washing could be associated with better overall performance. The Application C has confirmed its effectiveness. Proving its advantage among other applications, the Application C can be integrated into the clinical environment
- 2022
Dataset of Annotated Virtual Detection Line for Road Traffic Monitoring(atveras jaunā logā)
Ivars Namatēvs, Roberts Kadiķis, Anatolijs Zencovs, Laura Leja, Artis Dobrājs
Data, Sēj. 7 (4), lpp. 40, MDPI
Kopsavilkums
Monitoring, detection, and control of traffic is a serious problem in many cities and on roads around the world and poses a problem for effective and safe control and management of pedestrians with edge devices. Systems using the computer vision approach must ensure the safety of citizens and minimize the risk of traffic collisions. This approach is well suited for multiple object detection by automatic video surveillance cameras on roads, highways, and pedestrian walkways. A new Annotated Virtual Detection Line (AVDL) dataset is presented for multiple object detection, consisting of 74,108 data files and 74,108 manually annotated files divided into six classes: Vehicles, Trucks, Pedestrians, Bicycles, Motorcycles, and Scooters from the video. The data were captured from real road scenes using 50 video cameras from the leading video camera manufacturers at different road locations and under different meteorological conditions. The AVDL dataset consists of two directories, the Data directory and the Labels directory. Both directories provide the data as NumPy arrays. The dataset can be used to train and test deep neural network models for traffic and pedestrian detection, recognition, and counting.
Scopus procentile: 73 · SNIP: 1.128
- 2022
Interpretability versus Explainability: Classification for Understanding Deep Learning Systems and Models(atveras jaunā logā)
Ivars Namatēvs, Kaspars Sudars, Artis Dobrājs
Computer Assisted Methods in Engineering and Science, Sēj. 29 (4), lpp. 297 - 356, IPPT PAN
Kopsavilkums
The techniques of explainability and interpretability are not alternatives for many real-world problems, as recent studies often suggest. Interpretable machine learning is not a subset of explainable artificial intelligence or vice versa. While the former aims to build glass-box predictive models, the latter seeks to understand a black box using an explanatory model, a surrogate model, an attribution approach, relevance importance, or other statistics. There is concern that definitions, approaches, and methods do not match, leading to the inconsistent classification of deep learning systems and models for interpretation and explanation. In this paper, we attempt to systematically evaluate and classify the various basic methods of interpretability and explainability used in the field of deep learning. One goal of this paper is to provide specific definitions for interpretability and explainability in Deep Learning. Another goal is to spell out the various research methods for interpretability and explainability through the lens of the literature to create a systematic classifier for interpretability and explainability in deep learning. We present a classifierthat summarizes the basic techniques and methods of explainability and interpretability models. The evaluation of the classifier provides insights into the challenges of developing a complete and unified deep learning framework for interpretability and explainability concepts, approaches, and techniques.
Indeksēts Scopus / WoS · Scopus procentile: 73 · SNIP: 1
- 2022
QuinceSet: Dataset of annotated Japanese quince images for object detection(atveras jaunā logā)
Edīte Kaufmane, Kaspars Sudars, Ivars Namatēvs, Ieva Kalniņa, Jānis Judvaitis, Rihards Balašs, Sarmīte Strautiņa
Data in Brief, Sēj. 42, lpp. 108332, Elsevier
Kopsavilkums
With long-term changes in temperature and weather patterns, ecologically adaptable fruit varieties are becoming increasingly important in agriculture. For selection of candidate cultivars in fruit breeding or for yield predictions, fruit set characteristics at different growth stages need to be described and evaluated, which is largely done visually. This is a time-consuming and labor-intensive process that also requires sufficient expert knowledge. The annotated dataset for Japanese quince - QuinceSet - consists of images of Japanese quince (Chaenomeles japonica) fruits taken at two phenological developmental stages and annotated for detection and phenotyping. First, after flowering, when the second fruit fall is over and the fruits have reached 30-50% of their final size, and second, at the ripening stage of quince, just before the fruits are yielded. Both stages of quince images classified as unripe and ripe were annotated using ground truth ROI and presented in YOLO format. The dataset contains 1515 high-resolution RGB .jpg images with the same number of annotated .txt files. Images in the dataset were manually annotated using LabelImg software. A total of 17,171 annotations were provided by the experts. The images were acquired on site at the Institute of Horticulture in Dobele, Latvia. Homogenization of the images was performed under different weather conditions, at different times of the day, and from different capturing angles. The dataset contains both fully visible quinces and quinces partially obscured by leaves. Care was also taken to ensure that the foreground, which contains the leaves has adequate brightness with minimal shadows, while the background is darker. The presented dataset will allow to increase the efficiency of the breeding process and yield estimation, to identify and phenotype quinces more reliably, and may also be useful for breeding other crops.
Indeksēts Scopus / WoS · Scopus procentile: 72 · SNIP: 0.106
- 2022
YOLOv5 Deep Neural Network for Quince and Raspberry Detection on RGB Images(atveras jaunā logā)
Kaspars Sudars, Ivars Namatēvs, Jānis Judvaitis, Rihards Balašs, Artūrs Ņikuļins, Astile Peter, Sarmīte Strautiņa, Edīte Kaufmane, Ieva Kalniņa
2022 Workshop on Microwave Theory and Techniques in Wireless Communications (MTTW), IEEE
Kopsavilkums
Object detection based on deep learning can be widely used in all kinds of agricultural applications. In this paper, we present a deep neural network (DNN) model for quince and raspberry detection on RGB images. The trained DNN model is based on YOLOv5 architecture and it has 7 berry classes related to the berry development stage. YOLOv5 provides sufficiently good performance and precision trade-off. It is useful in the process of quince and raspberry phenotyping for the agriculture experts, where the yield and berry size parameters have to be estimated. Using our DNN model we have shown that it is possible to achieve a mean Average Precision close to 80.9 % and in some cases (Average Precision) close to 95 % for some classes. The DNN model is trained on labeled data gathered during the AKFEN project. The developed raspberry and quince detector is freely available at the GIT repository [1]. Further, the research on Sensor Networks, Wireless Systems, 3D point cloud processing and multi-spectral image processing has to be carried out leading to high-throughput phenotyping.
Indeksēts Scopus / WoS
- 2022
The Definitive Guide to Actual Sensor Network Deployments in Research Studies from 2013–2017: A Systematic Review(atveras jaunā logā)
Janis Judvaitis, Valters Abolins, Artis Mednis, Rihards Balass, Krisjanis Nesenbergs
Journal of Sensor and Actuator Networks, Sēj. 11 (4), lpp. 68, MDPI
Kopsavilkums
The research community has been working on sensor networks for more that seven decades and altogether more than a million research articles on sensor networks have been published, with this number growing every day. In this article, we try to provide a thorough and complete systematic review on the sensor network field resulting from representative subset of more than 3000 articles which include actual sensor network deployments. After a thorough analysis of these data, we provide a definitive overview of the trends in sensor network deployment described in published research articles
Scopus procentile: 93 · SNIP: 1.411
- 2022
Available Wireless Sensor Network and Internet of Things testbed facilities: dataset(atveras jaunā logā)
Janis Judvaitis, Valters Abolins, Amr Elkenawy, Kaspars Ozols
Open Research Europe, Sēj. 2 (127), lpp. 127, F1000 Research Limited
Kopsavilkums
The availability of data is an important aspect of any research as it determines the likelihood of the study's commencement, completion, and success. The Internet of Things and Wireless Sensor Networks technologies have been attracting a huge amount of researchers for more than two decades, without having a consolidated or unified source, identifying and describing available Internet of Things and Wireless sensor network testbed facilities. In this paper, a dataset including 41 distinct testbed facilities is described. These testbed facilities are classified according to their key features such as Device Under Test (DUT) type, mobility, access level, facility count, connection/interaction interfaces along with other criteria. The systematic review process resulting in the gathered data set consisted of three filtering phases applied to relevant articles published between the years 2011 and 2021 as obtained from the Web of Science and SCOPUS databases.
Indeksēts Scopus / WoS · Scopus procentile: 45
- 2022
Mobile IoT-Edge-Cloud Continuum Based and DevOps Enabled Software Framework(atveras jaunā logā)
Janis Judvaitis, Rihards Balass, Modris Greitans
Journal of Sensor and Actuator Networks, Sēj. 10 (4), lpp. 62, MDPI
Kopsavilkums
This research aims to provide a high-level software framework for IoT-Edge-Cloud computational continuum-based applications with support for mobile IoT and DevOps integration utilizing the Edge computing paradigms. This is achieved by dividing the system in a modular fashion and providing a loosely coupled service and module descriptions for usage in the respective system layers for flexible and yet trustworthy implementation. The article describes the software architecture for a DevOps-enabled Edge computing solution in the IoT-Edge-Cloud computational continuum with the support for flexible and mobile IoT solutions. The proposed framework is validated on an intelligent transport system use case in the rolling stock domain and showcases the improvements gained by using the proposed IoT-Edge-Cloud continuum framework.
Scopus procentile: 93 · SNIP: 1.411
- 2022
Intelligent Transport System: The Indra Use Case(atveras jaunā logā)
Francisco Parrilla, Sergio Jiménez Gómez, Modris Greitans, Janis Judvaitis
DevOps for Trustworthy Smart IoT Systems, Boston-Delft: now publishers
Kopsavilkums
This book reports on the ENACT project work to empower the development and operation as well as the continuous and agile evolution of Smart IoT Systems.The overall goal of the ENACT project was to provide a novel set of solutions to enable DevOps in the realm of trustworthy Smart IoT Systems. Smart IoT Systems are complex systems involving not only sensors but also actuators with control loops distributed all across the IoT, Edge and Cloud infrastructure.
Indeksēts Scopus / WoS
- 2022
Demonstration of Polarization Optical-Time-Domain Reflectometer for Monitoring of Optical Fiber Lines(atveras jaunā logā)
Janis Braunfelds, Sandis Spolitis, Lilita Gegere, Dmitrijs Pikulins, Vsevolods Stepanovs, and Andis Supe
2022 Workshop on Microwave Theory and Techniques in Wireless Communications (MTTW), IEEE
Kopsavilkums
Monitoring of optical signal polarization state in optical fiber networks can provide a variety of information for infrastructure owners. There is a rather limited choice of available techniques for monitoring the change in a polarization state of an optical signal over the transmission line. Mostly, as the typical measurement, the total optical line polarization mode dispersion (PMD) is determined, or the instantaneous change in polarization state at the line input or output is measured. But that does not allow determining the location and cause of sudden polarization changes in an optical fiber. The authors propose a polarization monitoring technique using a traditional commercially-used optical time-domain reflectometer (OTDR) with additional circuitry to generate polarization-dependent measurement traces. Such a technique can potentially be used for continuously tracking a physical impact on an optical fiber link.
- 2022
Convolutional Neural Network-based Image Restoration (CNNIR)(atveras jaunā logā)
Zheng-Jie Huang, Wei-Hao Lu, Brijesh Patel, Po-Yan Chiu, Tz-Yu Yang, Hao Jian Tong, Vytautas Bučinskas, Modris Greitans, Po Ting Lin
2022 18th IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications (MESA), IEEE
Kopsavilkums
In this era of automation, image processing is an indispensable part of computer vision. Many computer vision approaches in the industry depend on a relatively bright environment. Under low light source conditions, the distribution of image information is too concentrated in specific intensity ranges due to the color factor of the subject itself, resulting in noise and contrast loss. Enhancing contrast is a crucial step in improving the quality of the image and showing visible details. This study proposes a method based on a convolutional neural network (CNN), using the pixel difference between paired images, called a motion matrix, as an annotation for low-contrast images. The image's motion vector is predicted after the neural network model has been trained to produce the low-contrast enhanced image. Then, the proposed model is compared with the other methods.
Indeksēts Scopus / WoS
- 2022
An approach for assessment of concrete deterioration by surface waves(atveras jaunā logā)
A. Tatarinov, A. Sisojevs, A. Chaplinska, G. Shahmenko, V. Kurtenoks
Procedia Structural Integrity, Sēj. 37, lpp. 453-461, Elsevier
Kopsavilkums
The aging of concrete infrastructure facilities is a global problem. External and internal harmful factors such as water ingress, freezing and thawing, fire, chemical attack and forces lead to concrete degradation, frequently progressing from its surface. It is important to know how deep the deterioration process is expanded from the surface into the depth of the concrete bulk and how much the concrete degraded. The purpose of the study was to demonstrate the feasibility of an approach based on the application of pattern recognition methods to the analysis of ultrasonic signals to assess two variable factors of interest - the material degradation degree and the thickness of the conditionally weak surface layer on the strong concrete underlay. A series of specimens was made according to the 2-factor grid experimental design in order to build a mathematical model for the evaluation. Ultrasonic signals were acquired in the surface transmission at 2 frequencies, 50 to 100 kHz by stepped profiling of the specimens’ surface. 2D spatiotemporal waveform matrices acquired by surface profiling, where the coordinates were the propagation distance and ultrasonic time, served as the material for processing. The approach based on the methods of pattern recognition included steps of transformation the signals by the discrete Fourier transform, calculation of statistical criteria and creation of decision rules as bilinear mathematical approximation surfaces of the criteria. Verification of the method by substituting data on a control series of specimens into the mathematical model built on the grid of specimens of a training series allowed a realistic determination of the both factors of interest for a specimen from the control series. However, the use of only formal statistical criteria was found not completely sufficient for reliable recognition in all control cases. Further improvement of recognition can apply additional decision rules based on wave propagation in the time domain and refining of the existing decision rules by implementation of weights of influence.
Indeksēts Scopus / WoS · Scopus procentile: 41 · SNIP: 0.559
- 2022
Non-Destructive Quality Control of the Adhesive Rigid Timber-to-Concrete Connection in TCC Structures(atveras jaunā logā)
Buka-Vaivade K, Gaile L, Serdjuks D, Tatarinovs A, Pakrastins L
Buildings, Sēj. 12 (12), lpp. 2151, MDPI
Kopsavilkums
One of the limitations of using glued connections in practice is related to the need for connection quality control. Still, the need for the non-destructive quality control of finished products to determine the compliance of the developed structure with the designed one still exists. Considering the small amount of research on timber–concrete composites with glued connections, there is a lack of research on non-destructive methods for the quality control of rigid connections in timber–concrete composite structures. During the literature analysis, no information was found on the possibilities of testing the quality of the rigid timber-to-concrete connection. Therefore, two well-known methods—operational modal analysis and ultrasonic testing—were tested to verify the possibilities of applying these methods in determining defects in the rigid glued connection between the concrete and timber layers in the timber–concrete composite structures. A series of small-scale specimens produced by the stone chips method with and without artificially made defects in the timber-to-concrete adhesive connection was tested by both methods. Operational modal analysis shows significant changes in mode shape, frequency values, and spectral density diagrams. Despite the sufficiently large reflection of the ultrasonic signal on the timber and concrete boundary, the transmitted signal is sufficient to perform local ultrasonic tests for detecting defects in the adhesive connection. Thus, it is concluded that the principles of both methods can be applied in practice, and further research is needed to develop testing technology.
Indeksēts Scopus / WoS · Scopus procentile: 90 · SNIP: 1.372
- 2022
Extracting guided wave characteristics of bone phantoms from ultrasonometric data for osteoporosis diagnosis(atveras jaunā logā)
E. V. Glushkov, N. V. Glushkova, O. A. Ermolenko and A. M. Tatarinov.
2022 Days on Diffraction (DD), IEEE
Kopsavilkums
This work is a continuation of research presented at the last year DD meeting and partly published in [1], [2] and papers cited therein. Computer models for analytically based simulation of ultrasonic guided wave (GW) propagation in multilayer bone phantoms have been developed and used for systematic numerical analysis of GW propagation depending on phantoms' structure aimed to identify signs predicting osteoporosis. Now we have focused on developing reliable methods of processing experimental data to extract GW characteristics that can be used in the objective functions constructed for the inverse problem solution. Minimization of the objective functions in the space of waveguide's elastic constants and thicknesses yields its effective material parameters indicating a change in the strength properties.
Indeksēts Scopus / WoS
- 2022
Non-model vibration analysis method for health monitoring of structural joints(atveras jaunā logā)
D. Serdjuks, Viktors Kurtenoks, Aleksejs Tatarinovs, Karina Buka-Vaivade, Vjaceslavs Lapkovskis, Viktors Mironovs, Andrejs Podkoritovs, Kirils Topcijs
Procedia Structural Integrity, Sēj. 35, lpp. 555-562, Elsevier
Kopsavilkums
Structural health of joints plays an important role in the behavior of building structures. Structural Health Monitoring (SHM) methods make it possible to investigate and keep under control the behavior of structural joints during the entire service life of buildings. Choice of SHM methods depends on the material of the joined structural members, its loading case, joint’s structure and stiffness. Since the beginning of the 21st century, many SHM methods have been developed to investigate the behavior of structures using vibration analysis. Vibration analysis methods can identify possible damages of structures or structural joints due to changes in frequency response, modal shape and damping. Methods based on vibration analysis are classified into two categories - methods using ready-made mathematical models and methods that do not employ models. Current paper discusses various vibration analysis methods for assessing a structural health of joints. Particular attention is paid to non-model based methods, as it avoid the need for complex mathematical representations of models for specific investigation objects. In the context of the development of vibration analysis methods for building structures, the authors propose the concept of a 3D coaxial acceleration correlations to study the joints behavior in three-dimensional space.
Indeksēts Scopus / WoS · Scopus procentile: 41 · SNIP: 0.559
- 2022
Methods for Explaining CNN-Based BCI: A Survey of Recent Application(atveras jaunā logā)
Maksims Ivanovs; Beate Banga; Valters Abolins; Krisjanis Nesenbergs
2022 IEEE 16th International Scientific Conference on Informatics (Informatics), IEEE
Kopsavilkums
Convolutional neural networks (CNN) have achieved state-of-the-art results in many Brain-Computer Interface (BCI) tasks, yet their applications in real-world scenarios and attempts at further optimizing them may be hindered by their non-transparent, black box-like nature. While there has been ex-tensive research on the intersection of the fields of explainable artificial intelligence (AI) and computer vision on explaining CNN for image classification, it is an open question how commonly the methods for explaining CNNs are used when CNNs are a part of a BCI setup. In the present study, we survey BCI studies from 2020 to 2022 that deploy CNNs to find out how many of them use explainable AI methods for better understanding of CNNs and which such methods are used in particular. Our findings are that explainable AI methods were used in 13.7 percent of the surveyed publications, and the majority of the studies in which these methods were used employed the t-distributed stochastic neighbour embedding (t-SNE) method.
Indeksēts Scopus / WoS
- 2022
WearSec: Towards Automated Security Evaluation of Wireless Wearable Devices(atveras jaunā logā)
Bernhards Blumbergs, Ēriks Dobelis, Pēteris Paikens, Krišjānis Nesenbergs, Kirils Solovjovs and Artis Rušiņš
Lecture Notes in Computer Science / Nordic Conference on Secure IT Systems, Sēj. 13700, Springer Nature
Kopsavilkums
Wearable devices are becoming more prevalent in the daily life of society, ranging from smartwatches, and fitness bracelets to accessories and headphones. These devices, both from their hardware manufacturing and wireless firmware development perspectives may possess drawbacks. In recent years security researchers have uncovered a series of vulnerabilities. In this paper we introduce the concept and describe the key ideas towards the development of an automated security evaluation prototype for wireless wearable devices using device fingerprinting, as well as passive and active vulnerability identification. Furthermore we describe the technical approaches, challenges, and implementation choices we faced while developing the first stages of the prototype for this concept and handling full-spectrum Bluetooth analysis with software-defined radio.
Indeksēts Scopus / WoS · Scopus procentile: 50 · SNIP: 0.534
- 2022
Response Surface Method for Optimization of Synchronous Reluctance Motor Rotor
Orlova S., Pugachov V., Auzins J., Rassolkin A., Vaimann T.
Machines 2022