IFIP Open Access Digital Library
Not a member yet
22614 research outputs found
Sort by
Optimizing Hyperspectral Image Classification Through Swin Transformer Integration and CNN Feature Extraction
No full textPart 2: Data AnalyticsInternational audienceDue to the multidimensional structure of hyperspectral imaging, deep learning methods—in particular, vision transformer models—have been incorporated into classification systems. This article presents a novel use of Swin Transformer Network for hyperspectral image classification. The three layers that make up the suggested technique extract, embed, and merge image patches to provide a framework for classification. Our method uses patch-extraction, patch-embedding, and patch-merging approaches to improve feature representation. Hyperspectral features are preserved when window size, window shift, focus head, and input dimensions are taken into account. We demonstrate the effectiveness of our suggested method with a testing accuracy of 99.97% on hyperspectral datasets such as the Salinas public HSI dataset. The superiority of our approach is shown by comparing it with competing dimensionality reduction strategies. This work provides encouraging progress in the use of deep learning methods for classification of hyperspectral data
Deception Detection Using Random Forest and KNN
No full textPart 2: Data AnalyticsInternational audienceEvery person lies to try and escape or avoid some situation. Lying is considered to be a form of deceit and can have a very negative impact in today’s judicial world. Criminals, employees, shoplifters and many other people lie to get around the punishment. These must be detected so that innocent people are not put in adverse situations and also to catch real criminals. Conventional methods for detecting dishonesty entail the use of scientific methodologies to examine physiological signs, transcripts, as well as visual and audio data. Nevertheless, there has been a dearth of exploration into the utilization of modalities like EEG data. Despite years of research, scientists have consistently demonstrated that people’s ability to discern deception is no better than random chance. Precision in detecting deception holds paramount importance for law enforcement authorities. In the proposed model, brain waves and audio cues will be combined to enhance the existing deception detection. Existing datasets like ‘Bag-of-Lies’ will be utilized. This system is poised to make a substantial contribution to the Police Department, Criminal Investigation Department, and Judicial System by reducing the burden on law enforcement officers and aiding in the accurate identification of the actual perpetrator
Research on Factor Support Vector Multi-classification Algorithm Based on Factor Space Theory
No full textPart 2: Causal ReasoningInternational audienceIn order to solve the accuracy and complexity of multiple classification algorithms, on the basis of factor space theory, the divisibility measurement condition and the construction condition of binary tree are defined. On the basis of balanced binary tree, combining the relationship between class spacing and sample circle radius, a factor support vector multiple classification algorithm(M-FSV) is proposed by using recursion idea and the principle of “easy classification first”. Experiments are done with one-to-one support vector machine and balanced binary tree support vector machine algorithm, and experimental comparison is made. The results of 8 data sets in UCI database show that the training time of M-FSV is less than that of SVM, and the algorithm accuracy is higher than that of SVM. The research results expand the theory and application of factor space, and provide a new idea and simple method for classification problems in machine learning
Research on Improvement of Sweeping Learning Chain Algorithm Based on Factor Space Theory
No full textPart 1: Machine LearningInternational audienceIn order to improve the classification accuracy of sweeping learning chain (SLC) algorithm in factor space and solve the problem that SLC is error- prone to classify samples in mixed domain, this paper proposes sweeping learning chain-K-Nearest Neighbor (SLC-KNN) algorithm. When SLC encounters the problem of undivisible data, KNN is used to classify the samples to be tested that fall into the mixed domain. It not only solves the problem of undivisible data encountered by the SLC, but also reduces the amount of computation and storage of KNN algorithm. And extended to multi-classification problem, proposed BT- SLC-KNN multi-classification algorithm. According to the maximum class center distance, firstly the algorithm makes the two most easy to separate classes separate. And defines the Class center distance on sweeping vector, a normal binary tree is generated step by step by comparing the distance of the remaining categories on the sweeping vector of the two classes with the furthest distance. It reduces the time complexity of merge sweeping learning chain (MSLC) algorithm and improves the multi-classification accuracy. Finally, the experiments are carried out on UCI data sets, and the results show that the two algorithms are feasible and effective
Detection of Depression in EEG Signals Based on Convolutional Transformer and Adaptive Transfer Learning
No full textPart 1: Machine LearningInternational audienceElectroencephalography (EEG) signals provide an objective reflection of the inner workings of the brain, making them a promising tool for the diagnosis of depression. However, the classification of EEG signals for depression is severely affected by individual differences among subjects, complex intrinsic properties, and low Signal-to-Noise Ratio (SNR), which limits the classification accuracy. Additionally, traditional convolutional neural networks extract local features but fail to capture long-term dependencies in EEG decoding. To address the aforementioned issues, we introduce an adaptive transfer learning method based on a convolutional transformer model for depression detection. The experimental results demonstrate the effectiveness of the proposed model on the public MODMA dataset and EDRA dataset. The results indicate that the MODMA and EDRA datasets exhibit optimal accuracies of 100% and 98.61%, respectively, outperforming some state-of-the-art depression identification methods. Our findings provide new perspectives on the recognition of depression, which could be used as an assisted diagnostic tool in the future
Difference-Enhanced Learning of the Deep Semantic Segmentation Networks for First Break Picking
No full textPart 1: Machine LearningInternational audienceThe precise estimation of seismic arrival times, commonly referred to as first-break picking, is a critical problem in seismic research due to its important role in various seismological applications such as statics correction processing. In recent years, there have been several deep learning algorithms designed specifically for 2D seismic arrival time picking. A widely used approach is to treat the 2D arrival picking problem as a 2D image segmentation problem and employ a deep semantic segmentation model for end-to-end first break picking. However, the first break mask generated from this method often fails to meet the uniqueness of first arrival time according to certain noises. In order to alleviate this problem, we propose a difference-enhanced learning method of the deep semantic segmentation network for the first break picking problem by designing a new kind of loss function, which actually improves the quality of mask generation and arrival time accuracy. It is demonstrated by extensive experiments on a real seismic dataset that our proposed difference-enhanced learning method is effective and outperforms the conventional learning methods for deep semantic segmentation models on the estimation of seismic arrival times
Sentiment Analysis for Stock Prediction Using Mass Media Sources
No full textPart 1: Applications of AI/ML in Natural Language ProcessingInternational audienceSentiment Analysis for Stock Prediction Using Mass Media Sources" introduces a groundbreaking approach to forecasting stock movements by harnessing sentiment analysis applied to economic news gathered from a diverse range of mass media sources. The project encompasses the creation of a comprehensive system that systematically scrapes economic news websites to gather information relevant to specific companies. This data is then meticulously analyzed to discern the sentiments expressed in the news articles. Subsequently, cutting-edge Machine learning techniques are put to use to anticipate potential fluctuations in the stock prices of the target companies. This holistic approach capitalizes on the amalgamation of Natural language processing which combined with machine learning techniques, and real-time news data, delivering invaluable insights for both investors and traders
Design and Analysis of Structural Health Monitoring System for the Diagnosis of Morphological Deformities of Bolted Structures
No full textPart 2: Applications of AI/ML in Image ProcessingInternational audienceA procedure called structural health monitoring (SHM) aims to deliver accurate and timely information on a structure’s performance. The primary concern of the State agencies participating in the many elements of their inquiry, planning, design, building, operation, and maintenance is the safety of the numerous engineering structures in our nation. Due to these developments, there is an increasing need to cut maintenance costs and create a safer environment by preventing structural breakdowns. The degradation of engineering structures brought on by time, incorrect maintenance, and other internal and external environmental elements has had a substantial impact on the quality of structures. Therefore, it is essential to establish SHM Systems for the continuous or periodic long-term monitoring of vital parameters in order to assure safety and reliability. The integrity of bolted joints in steel structures is a growing problem since it is possible for bolts to self-loosen as a result of various external environmental conditions. Using a Single-Ended Primary-Inductor Converter (SEPIC), a DC/DC buck-boost converter topology, a sensor-based bolt looseness detection and reporting system for railway fishplate joints has been proposed in this work. Setting the duty cycle of the control switch to provide a constant voltage output requires a PI Controller. An ultrasonic sensor was used to detect looseness, and an algorithm for identifying and evaluating structural anomalies has been proposed. The conceivable future advancement designs that can be coordinated with this framework are likewise shown in this paper
Multi-camera Enhanced Real-Time Content-Aware Vehicle Detection
No full textPart 1: Applications of AI/ML in Natural Language ProcessingInternational audienceThis paper introduces an open-source solution for real-time object awareness in the context of intelligent vehicles, designed to seamlessly adapt to simulation environments. The system transcends traditional vehicle detection, offering a comprehensive framework that includes object classification, precise location, and rotation estimation. Leveraging a multi-camera setup, our approach provides a comprehensive understanding of the surrounding environment, extending beyond vehicle recognition to include various other objects. What sets our solution apart is its capability to efficiently reformat and transition to simulation environments like Blender, enabling integration and testing within virtual contexts. The core components of our system encompass data collection, preprocessing, deep learning model selection, Use of Look up table, model training, efficient inference, and post-processing techniques. Designed for open-source collaboration, this solution is positioned for continuous improvement, adaptation to evolving needs, and addressing emerging challenges in the field of intelligent transportation and related domains. This paper represents a foundational step towards establishing an accessible and adaptable real-time object awareness system, encouraging innovation and research in the realm of intelligent vehicles and their applications, both in the real world and in simulated environments
Offensive Language Detection on Telugu Language
No full textPart 1: Applications of AI/ML in Natural Language ProcessingInternational audienceIn the present world a lot of data is generated via twitter, Instagram, WhatsApp etc. in different languages. It is important and necessary task to detect the offensive language among those data to create healthy and good environment among people. And it is even highly challenging task to identify offensive language in low resource languages due to less availability of the classified datasets. This paper aims to detecting of offensive language on a low resource language Telugu. To identify solution for this problem different types of ML models and DL models are used. Based on accuracy of the different models we are going to choose a model. To split the data and train the data and to test the data we have been using stratified k fold cross validation which is an efficient way to split the data and to increase model’s ability to perform better. From basis of this experiments, we can have a model to detect offensive language in Telugu and it must be considered as a small step for future models to work on it