the real area where it is well researched is Human Activity recognition in the field of Computer Vision. Computer vision is one of the well-explored areas of interest in the proper processing and analysis of visual in...
详细信息
Micro-expressions, which are brief facial movements reflecting an individual's emotions and psychological state, are crucial in fields like psychological counseling and trust evaluation. However, they are challeng...
详细信息
ISBN:
(纸本)9798350375084;9798350375077
Micro-expressions, which are brief facial movements reflecting an individual's emotions and psychological state, are crucial in fields like psychological counseling and trust evaluation. However, they are challenging to capture, and traditional methods struggle with effective recognition. While neural networks are a primary tool for this task, they are limited by their large number of parameters and slow processing speeds. Transformer architectures have shown promise in micro-expression recognition but face similar challenges. To address these issues, an improved HTNet model, called Lightweight HTNet (LWHTNet), has been proposed. Specifically, by using a Separable Self-Attention mechanism to replace the multi-head attention mechanism in Transformer layers and incorporating depthwise separable convolutions in the block aggregation layers, the model reduces the number of training parameters. In experiments, LWHTNet demonstrated strong performance on composite datasets, achieving an Unweighted F1 Score (UF1) of 0.8498 and an Unweighted Average Recall (UAR) of 0.8307. Additionally, the optimized model showed significant improvements in inference speed and parameter efficiency.
As the demand for deploying machinelearning (ML) and Deep learning (DL) models at the network's edge continues to grow, it becomes imperative to evaluate and compare the performance of various edge platforms. thi...
详细信息
ISBN:
(纸本)9783031821493;9783031821509
As the demand for deploying machinelearning (ML) and Deep learning (DL) models at the network's edge continues to grow, it becomes imperative to evaluate and compare the performance of various edge platforms. this study introduces an innovative edge-based pipeline for analyzing home energy and environmental data and addressing challenges in managing imbalanced label distributions. the comparison between random under-sampling and over-sampling strategies reveals the superiority of over-sampling, significantly improving accuracy and f1-score. More specifically, over-sampling using the Decision Tree (DT) technique enhanced accuracy and f1-score by 4.97% and 9.24%, respectively, when compared to the original results. the paper expands its contribution by conducting a thorough performance evaluation of various edge platforms, emphasizing critical metrics. the evaluated edge platforms include a Raspberry Pi 4b+, Jetson Nano Developer Kit, Odroid XU4, and Coral Dev Board. the study conducted an exhaustive comparison of the runtime of these prominent edge platforms when running ML/DL algorithms for tabular and image datasets during the testing phase. the Jetson Nano, particularly with its built-in GPU, demonstrates superior performance in ML/DL tasks, showcasing accelerated processing times for Convolutional Neural Network (CNN)-based algorithms. these findings emphasize the efficacy of the Jetson Nano in handling computationally intensive tasks for edge-based applications.
Cybersecurity remains a grand societal challenge. Large and constantly changing attack surfaces are non-trivial to protect against malicious actors. Entities like the United States and the European Union have recently...
详细信息
ISBN:
(纸本)9798400704901
Cybersecurity remains a grand societal challenge. Large and constantly changing attack surfaces are non-trivial to protect against malicious actors. Entities like the United States and the European Union have recently emphasized the value of Artificial Intelligence (AI) for advancing cybersecurity. For example, the National Science Foundation has called for AI systems that can enhance cyber threat intelligence, detect new and evolving threats, and analyze massive troves of cybersecurity data. the 4(th) Workshop on Artificial Intelligence-enabled Cybersecurity Analytics (co-located with ACM KDD) sought to make significant and novel contributions within these relevant topics. Submissions were reviewed by highly qualified AI for cybersecurity researchers and practitioners spanning academia and private industry firms.
the accurate identification of script styles in historical manuscripts is crucial for gaining insights into their historical context and significance. this study proposes an integrated system that combines manual and ...
详细信息
ISBN:
(纸本)9783031821554;9783031821561
the accurate identification of script styles in historical manuscripts is crucial for gaining insights into their historical context and significance. this study proposes an integrated system that combines manual and machinelearning features to effectively identify script styles within manuscripts, utilizing the ClAMM dataset. the system comprises three primary steps: preprocessing the dataset using denoising and binarization techniques, extracting manual features using the Harris detector, and performing script classification using pretrained CNN models. By merging manual feature engineering with advanced deep learning techniques, our system showcases its ability to accurately recognize script styles competing with state-of-the-art methods with accuracy of 89.2%. these results not only validate the effectiveness of our approach but also contribute significantly to the broader advancement of script classification in historical manuscript analysis.
Withthe rapid development of information technology, the widespread dissemination of long text data in cyberspace has brought challenges to personal privacy and information security. this study aims to develop an int...
详细信息
It is essential that institutions use Educational datamining and Deep learning to recognize academically poor students and assist them by establishing various recommendation systems to improve their effectiveness. By...
详细信息
the proceedings contain 200 papers. the topics discussed include: heat-aware graph data placement strategy for NVM;parameters estimation of photovoltaic models via an improved differential evolution algorithm;study on...
ISBN:
(纸本)9798400708831
the proceedings contain 200 papers. the topics discussed include: heat-aware graph data placement strategy for NVM;parameters estimation of photovoltaic models via an improved differential evolution algorithm;study on the extraction of law enforcement relationships in administrative law enforcement instrument data;online algorithm for exploring a grid polygon with two robots;research on gesture recognition method by improving dung beetle algorithm to optimize BP neural network;an improved algorithm for frequent sequence patternmining based on PrefixSpan-ComplexPrefixSpan;machinelearning-based research on reserve prediction of natural-gas-hydrates;enhancing coal mine safety monitoring algorithm using graph computing techniques;reverse distillation support vector data description for unsupervised anomaly detection;and few-shot object counting model based on self-support matching and attention mechanism.
the intersection of computer science and social sciences leads an increasing problem: fraudulent job ads. their presence can have far-reaching negative ramifications that must be dealt with immediately in order to lim...
详细信息
ISBN:
(纸本)9798350386356;9798350386349
the intersection of computer science and social sciences leads an increasing problem: fraudulent job ads. their presence can have far-reaching negative ramifications that must be dealt with immediately in order to limit further harm to society. Due to the vast amounts of data accumulated worldwide, effective ways must be found for distinguishing legitimate job advertisements from fraudulent ones. this paper proposes a machinelearning approach designed to differentiate between fraudulent and non-fraudulent job postings, offering an automated tool to mitigate the proliferation of fraudulent job listings on the internet. the methodology relies on a range of machinelearning classification techniques to analyze web-based job postings, comparing the efficacy of different classifiers in identifying employment scams. the primary objective is to establish a robust model for the detection of fake job posts amidst the overwhelming volume of online listings. this research recommends employing various datamining and categorizing algorithms such as Decision Tree and Support Vector machine, Naive Bayes classifier, Random-forest classifier and Multilayer Perceptron to detect whether an advertisement is genuine. this study has utilized Kaggle data withthe proposed classifier enhancing 99.48% accuracy when classifying fake job ads. Additionally, this method provides techniques to detect employment scams to meet applicants' need to protect themselves against scammers.
the proceedings contain 39 papers. the special focus in this conference is on data Science, machinelearning and Applications. the topics include: Monitoring the Farming Conditions Using IoT;Design of OTFS Modulation ...
ISBN:
(纸本)9789819920570
the proceedings contain 39 papers. the special focus in this conference is on data Science, machinelearning and Applications. the topics include: Monitoring the Farming Conditions Using IoT;Design of OTFS Modulation by Superimposed Pilot-Based Channel Estimation and Embedded Pilot-Aided Estimation;Morphology and EMD-Based Patch-Wise Image Fusion;PSO-Based Evolutionary Image Segmentation System for Analysis of Fatty Liver Level recognition;blake–Zisserman Model of Segmentation Method for Low-Contrast and Piecewise Smooth Image;Performance Enhancement of Wireless Systems Using Hybrid RIS Technique;Creation of a Platform for Artisans to Promote their Product Using Blockchain as NFT;face recognition Based Home Security System to Detect Usual/Unusual person Using IoT;opinion mining-Based Fake Review Detection Using Deep learning Technique;automatic Detection and Cleaning of Manhole Blockages Using IoT;design and Implementation of a Smart Door Locking System with Automatic Appliance Switching;dynamic Game Difficulty Adjustment Based on Facial Emotion recognition;parallel Implementation of PageRank Based K-Means Clustering on a Multithreaded Architecture;ZACube-2 Mission Operations Analysis;a Novel Approach for Speech Emotion recognition with Facial Expression Analysis;effect of Environment on Students Performance through Orange Tool of datamining;phishing Email Mitigation Technique Using Back-Propagation Neural Network for Cyber Space;crime Visualization and Forecasting Using machinelearning;an Empirical Comparison of Classification machinelearning Models Using Medical datasets;Analysis of the SEER data set for Lung Cancer Diagnosis for Stage Classification and Survival Analysis;risk Prediction of Chronic Kidney Disease Using machinelearning Algorithms;sandalwood Tree Deduction Using Deep learning;phishing Attack Detection Using machinelearning.
暂无评论