Article
Computer Science, Information Systems
Ali Arshaghi, Mohsen Ashourian, Leila Ghabeli
Summary: Using machine vision and image processing methods plays an important role in identifying defects in agricultural products, particularly potatoes. Research has shown that applying image processing and artificial intelligence in agriculture can improve the accuracy of identifying and classifying pests and diseases. In this study, a convolution neural network (CNN) method was used to analyze five classes of potato diseases, and the results were compared with other methods. The findings demonstrate that the proposed deep learning method achieved higher accuracy compared to existing works.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Energy & Fuels
Nils Muller, Carsten Heinrich, Kai Heussen, Henrik W. Bindner
Summary: This study proposes a data processing pipeline to identify flexibility activation events in real-time, which can help distribution system operators monitor and control the operation of distribution networks. By detecting and verifying flexibility activations from electricity markets and DSOs, the awareness of critical grid situations can be increased and counteractions can be taken to prevent issues.
Article
Computer Science, Interdisciplinary Applications
You Peng, Birgit Braun, Casey McAlpin, Michael Broadway, Brenda Colegrove, Leo Chiang
Summary: Ensuring contamination-free polyethylene products is crucial for many applications. To automate the classification of contaminants in polymer production, a multi-class classifier was built using convolutional neural networks. The trained model achieved an accuracy of over 95% on the test set.
COMPUTERS & CHEMICAL ENGINEERING
(2022)
Article
Computer Science, Theory & Methods
William C. Sleeman, Rishabh Kapoor, Preetam Ghosh
Summary: This study proposes a new taxonomy for describing multimodal classification models, aiming to address challenges in the field such as inconsistent terminologies and architectural descriptions, as well as unresolved issues like big data, class imbalance, and instance-level difficulty. The paper presents examples of applying this taxonomy to existing models and offers a checklist for the clear and complete presentation of future models.
ACM COMPUTING SURVEYS
(2023)
Article
Computer Science, Interdisciplinary Applications
Yi Xing, Liyong Tong
Summary: In this work, a machine learning-assisted structural optimization (MLaSO) scheme is proposed to accelerate the computational speed of structural optimization. A new machine learning model is used to predict the update of the optimization quantity during the optimization process, eliminating the need for finite element analysis and sensitivity analysis. The MLaSO scheme can be easily integrated into different structural optimization methods and does not require additional training datasets.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2022)
Article
Computer Science, Artificial Intelligence
Lorraine Chambers, Mohamed Medhat Gaber
Summary: Convolutional Neural Networks (CNNs) can achieve state of the art results for visual recognition problems when the data distributions are the same between train and test sets and all test set classes are present in the training data. However, in the real world, where data evolves and new classes emerge, traditional neural networks fail to identify unknown classes. Open-Set Classification research field provides potential solutions for this problem. In this study, a system called DeepStreamOS is proposed, which combines deep neural network activations with a stream-based outlier detection method to quickly identify instances belonging to unknown classes. Experimental results demonstrate that DeepStreamOS outperforms other open-set classification methods in most scenarios and significantly improves the speed of classification.
PATTERN RECOGNITION LETTERS
(2022)
Article
Physics, Multidisciplinary
Milosz Gajowczyk, Janusz Szwabinski
Summary: In this study, deep residual networks were utilized to classify molecular trajectories in living cells, resulting in a model with higher accuracy, fewer parameters, shorter training time, reduced overfitting, and better generalization to unseen data compared to the initial network.
Article
Computer Science, Information Systems
Xiao Tan, Sheng Li, Hua Yan
Summary: Currently, although significant progress has been made in lane detection using deep learning methods in complex scenarios, there is still room for improvement in terms of real-time performance. The mainstream approach to improve real-time performance is row-wise classification, which compromises accuracy for speed. However, models using this approach often lack the ability to extract spatial contextual information, which hinders lane recognition. Inspired by Feature Pyramid Networks, we propose SIE-Net, a simple and lightweight framework based on row-wise classification. Our method fully extracts spatial information in the image, fusing semantic and spatial information from deep and shallow feature maps respectively. We also utilize dilated convolution and channel attention mechanism in the feature extraction process to extract global information and give more weight to lanes' structural information. Experimental results on two benchmark datasets, Tusimple and CULane, demonstrate the effectiveness of our proposed method in terms of accuracy and speed.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Jianxiong Tang, Jian-Huang Lai, Xiaohua Xie, Lingxiao Yang, Wei-Shi Zheng
Summary: This paper proposes a fast and memory-efficient Activation Consistency Coupled ANN-SNN (AC2AS) framework for training SNN in low-power environments. The framework utilizes a weight-shared architecture between ANN and SNN, as well as spiking mapping units, to achieve fast training and ensure activation consistency for SNN. Experimental results show that AC2AS-based models perform well on benchmark datasets and achieve comparable accuracy with reduced time steps, training time, GPU memory costs, and spike activities compared to the Spike-based BP model.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Software Engineering
Filipe Assuncao, Nuno Lourenco, Bernardete Ribeiro, Penousal Machado
Summary: This paper introduces a grammar-based general purpose framework for the automatic search and deployment of potentially Deep Artificial Neural Networks (DANNs). Fast-DENSER optimizes the topology, learning strategy, and required hyper-parameters simultaneously for tasks like object recognition. The code is developed and tested in Python3, and a simple example for the automatic search of CNNs for the Fashion-MNIST benchmark is provided.
Article
Astronomy & Astrophysics
B. S. Kronheim, M. P. Kuchera, H. B. Prosper, A. Karbo
Summary: Machine learning is used to model the parameter space of the phenomenological Minimal Supersymmetric Standard Model (pMSSM) to predict its results. The results demonstrate the potential for machine learning to accurately predict the outcomes of new physics theories, providing possibilities for further research.
Article
Computer Science, Artificial Intelligence
R. Arumuga Arun, S. Umamaheswari
Summary: This paper proposes a multi-crop disease detection model using a complete concatenated deep learning architecture. The architecture introduces complete concatenated blocks as core functional units and reduces model size through pruning. The best performing model achieves a classification accuracy of 98.14% and has a smaller model size.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Truong X. Tran, Ramazan S. Aygun
Summary: The study introduces trustable learning to avoid misclassification errors in machine learning applications, where the model will reject making a decision or defer it to a human expert when unable to classify accurately. By applying WisdomNet architecture, it is possible to develop a classifier with 0% misclassification error and reduce the classification error rate to 0% in various network architectures.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Engineering, Geological
Ioannis Farmakis, Paul-Mark DiFrancesco, D. Jean Hutchinson, Nicholas Vlachopoulos
Summary: This study explores the potential of integrating complex deep learning architectures into dynamic rockfall database population processes, aiming to relieve experts from manually classifying large amounts of change detection data. Utilizing deep neural networks based on PointNet and PointNet++ architectures, developed on a 5-year change detection database of over 8000 rock slope clusters for training, the models achieve high accuracy levels on the last two data acquisitions and a geologically different rockfall database.
ENGINEERING GEOLOGY
(2022)
Article
Environmental Sciences
Hassan Afzaal, Aitazaz A. Farooque, Arnold W. Schumann, Nazar Hussain, Andrew McKenzie-Gopsill, Travis Esau, Farhat Abbas, Bishnu Acharya
Summary: This study evaluated the potential of using machine vision combined with deep learning to identify early blight disease in potato production systems. By training CNN models, the study accurately classified the disease at various growth stages, with EfficientNet performing the best.
Article
Computer Science, Information Systems
Galina Kamyshova, Aleksey Osipov, Sergey Gataullin, Sergey Korchagin, Stefan Ignar, Timur Gataullin, Nadezhda Terekhova, Stanislav Suvorov
Summary: This article proposes a methodology for optimizing crop irrigation using a phytoindication system based on computer vision methods. The system, divided into three stages including image preprocessing, classification, and neural network training, achieves a high accuracy rate of 93% in plant identification. The system can process up to 100 plants per second, surpassing similar systems in performance.
Article
Mathematics
Igor Timofeev, Ekaterina Pleshakova, Elena Dogadina, Aleksey Osipov, Azret Kochkarov, Stefan Ignar, Stanislav Suvorov, Sergey Gataullin, Sergey Korchagin
Summary: This study investigates the suitability of the electroflotation coagulation method for extracting vegetable and milk proteins and proposes a technological scheme for protein extraction. The study explores the use of new electrolysis parameters and protein isoelectric state regulation to enhance extraction efficiency.
Article
Mathematics
Alisa Batmanova, Alexander Kuc, Vladimir Maksimenko, Andrey Savosenkov, Nikita Grigorev, Susanna Gordleeva, Victor Kazantsev, Sergey Korchagin, Alexander E. Hramov
Summary: This study trained an artificial neural network using 32 EEG channels to predict errors in perceptual decision-making. By transforming the 2D input data into a 1D feature vector through convolutional procedures, the model achieved high accuracy in predicting perceptual decision-making errors. The findings have significant implications for predicting and preventing human errors in brain-computer interfaces.
Article
Multidisciplinary Sciences
Junwei Zhang, Xin Kang, Yang Liu, Huawei Ma, Teng Li, Zhuo Ma, Sergey Gataullin
Summary: This paper proposes a secure and lightweight multi-party private intersection-sum scheme, SLMP-PIS, which maintains data privacy based on zero sharing and arithmetic sharing, and considers the privacy of associated values using symmetric encryption. The security analysis shows that our protocol is secure in the semi-honest security model, and the experimental results demonstrate its efficiency and feasibility, with a 22.98% increase in efficiency when the number of participants is five.
Article
Computer Science, Information Systems
Dmitry Tsapin, Kirill Pitelinskiy, Stanislav Suvorov, Aleksey Osipov, Ekaterina Pleshakova, Sergey Gataullin
Summary: This article discusses the trends in the introduction of industrial and logistics robots to ensure the safety of civilian facilities, the problems of increasing crime in the Russian Federation, and the decline in the identification of criminals. It also explores the use of BigData methods, specifically an ensemble of computer vision algorithms and convolutional neural networks, for timely detection of emergency situations in parking lots using mobile robots. The implementation of training on convolutional neural networks and the addition of a Squeeze-and-Excitation block (SE) improved the accuracy by 2-3%, reaching 88%, 91%, and 92% respectively. Comparison with the HOG-BoVW-BPNN method showed that DenseNet121 + SE achieved the same accuracy of 86% but with a 40% faster speed, making it a more attractive option for a car park computer vision system.
JOURNAL OF COMPUTER VIROLOGY AND HACKING TECHNIQUES
(2023)
Article
Computer Science, Information Systems
Alexey Osipov, Ekaterina Pleshakova, Yang Liu, Sergey Gataullin
Summary: This paper studies human behavior in stressful situations using machine learning methods, which is influenced by psychotype, socialization, and other factors. The study reveals that global mobile subscribers lost around $53 billion in 2022 due to phone fraud, with nearly half of them having spam blocking or caller ID apps installed. By narrowing down the target audience to males under the age of 44, who are at the highest risk of being deceived by scammers, the researchers were able to develop a modified neural network that can detect panic stupor states in real-time and respond to phone scammers during conversations with subscribers.
JOURNAL OF COMPUTER VIROLOGY AND HACKING TECHNIQUES
(2023)
Article
Computer Science, Information Systems
Alexey Osipov, Ekaterina Pleshakova, Artem Bykov, Oleg Kuzichkin, Dmitry Surzhik, Stanislav Suvorov, Sergey Gataullin
Summary: The purpose of this article is to develop an effective method for monitoring the state of the drill string and the bit in low-time delay mode. A experimental setup was created based on the phase-metric control method to continuously monitor the drilling process. By analyzing the electrical characteristics of the probing signal, the authors used deep learning methods to identify the state of the drill string and the bit. The WFT-2D-CapsNet method showed high accuracy in detecting transitions between rock layers and the condition of the bit.
Proceedings Paper
Computer Science, Cybernetics
A. A. Kositzyn, A. S. Bogomolov, A. F. Rezchikov, V. A. Kushnikov, V. A. Ivashchenko, J. V. Lazhauninkas, R. B. Nurgaziev, L. A. Sleptsova, E. V. Berdnova, S. A. Korchagin, D. V. Serdechnyy
Summary: This article investigates the problem of emergency combinations of events during the inspection of industrial facilities by unmanned aerial vehicles. A formal and meaningful formulation of the problem is proposed, and fault trees are used to represent the development of accidents from combinations of individually non-hazardous events. The fault trees include hardware and software malfunctions, operator errors, and environmental influences.
CYBERNETICS PERSPECTIVES IN SYSTEMS, VOL 3
(2022)
Article
Robotics
Oleg Krakhmalev, Sergey Gataullin, Eldar Boltachev, Sergey Korchagin, Ivan Blagoveshchensky, Kang Liang
Summary: This article introduces the concept of an automated system for harvesting apples. The system consists of a robotic complex mounted on a tractor cart, including an industrial robot and a packaging system with a container for fruit collection. The robot is equipped with a vacuum gripper and a vision system. A power supply generator, vacuum pump for the gripper, and equipment control system are also installed on the cart. The developed automated system has a high degree of reliability to meet field operation requirements.
Article
Computer Science, Information Systems
Aleksey Osipov, Vyacheslav Shumaev, Adam Ekielski, Timur Gataullin, Stanislav Suvorov, Sergey Mishurov, Sergey Gataullin
Summary: This study investigates the use of machine learning methods to detect mechanical damage in sugar beetroot crops for fine-tuning beet harvester units. The Agrifac HEXX TRAXX harvester with a computer vision system was utilized. Image processing and classification methods were applied to accurately identify damage in sugar beetroots.