IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Security in smart cities using YOLOv8 to detect lethal weapons
Get PDFThe increase in the illegal use of lethal weapons at a global level has reachedworrying figures, resulting in an increase in assaults and armed robberies. Based on the above, closed circuit television (CCTV) surveillance systems emerge as an alternative solution. Therefore, the use of artificial intelligence is explored in order to detect the presence of lethal weapons in images accurately. In this study, a convolutional neural network model YOLOv8 is trained. A database including 4104 images with the presence of lethal weapons is generated. The Google Colab platform is used for the training phase, since it provides us with a free graphic processing unit (GPU), and the YOLOv8x and YOLOv8n models are used for comparison. The results indicate an advantage when using the YOLOv8 models, and when comparing them with similar models already proposed in the studied literature, we can conclude that our model stands out with an accuracy of 89.56% in the detection of lethal weapons. In conclusion, we were able to obtain a model capable of detecting lethal weapons in CCTV images, in addition to being able to be used in applications that require real-time detection.
Parallel rapidly exploring random tree method for unmanned aerial vehicles autopilot development using graphics processing unit processing
Get PDFAutonomous air movement systems hold great potential for transforming various industries, making their development essential. Autopilot design involves advanced technologies like artificial intelligence, machine learning, and big data. This paper focuses on developing a parallel rapidly-exploring random tree (RRT) algorithm using compute unified device architecture (CUDA) technology for efficient processing on graphics processing units (GPUs). The study evaluates the algorithm's performance in automated trajectory planning for unmanned aerial vehicles (UAVs). Numerical experiments show that the parallel algorithm outperforms the sequential central processing unit (CPU)-based version, especially as task complexity and state space dimensions increase. In scenarios with numerous obstacles, the parallel algorithm maintains stable performance, making it well-suited for various applications. Comparisons with CPU-based methods highlight the advantages of GPU use, particularly in terms of speed and efficiency. Additionally, the performance of two GPU models, NVIDIA RTX 2070 and T4 is compared, with the T4 demonstrating superior performance for similar tasks. Future research should explore integrating multiple algorithms for a more comprehensive UAV autopilot system. The proposed approach stands out for its stability and practical applicability in real-world autopilot implementations
Schedule-free optimization of the transformers-based time series forecasting model
Get PDFThe task of time series forecasting is important for many scientific, technical, and applied fields, such as finance, economics, meteorology, medicine, transportation, and telecommunications. Existing methods, such as autoregressive models and moving average models, have their limitations, especially when working with non-stationary and seasonal data. In this work, the basic architecture of transformers was modified to solve time series forecasting problems. Additionally, state-of-the-art optimizers were investigated and experimentally compared, including AdamW, stochastic gradient descent (SGD), and new methods such as schedule-free SGD and schedule-free AdamW, to improve forecasting accuracy and the efficiency of the training procedure for the transformer architecture. Modeling was conducted on meteorological data that included seasonal time series. The accuracy evaluation of the optimization methods studied in this work was performed using a range of different performance indicators. The results showed that the new optimization methods significantly improve forecasting accuracy compared to the use of traditional optimizers
Enhancing accessibility with long short-term memory-based sign language detection systems
Get PDFIndividuals who are deaf or experience difficulties with hearing and speech predominantly rely on sign language as their medium to communicate, which is not universally comprehended leading to obstacles in achieving effective communication. Advances in deep learning technologies in recent years have enabled the development of systems intended to autonomously interpret gestures in sign language and translate them into spoken language. This paper introduces a system built on deep learning methodologies for recognizing sign language. It uses long short-term memory (LSTM) architecture to distinguish and classify hand gestures which are static and dynamic. The system is divided into three primary components, including dataset collection, neural network assessment, and sign detection component that encompasses hand gesture extraction and sign language classification. The module to extract hand gestures makes use of recurrent neural networks (RNNs) for the detection and tracking of hand movements in video sequences. Another RNN that is incorporated by classification module categorizes these gestures into established sign language classes. Upon evaluation on a custom dataset, the proposed system attains an accuracy rate of 99.42%, thus making it visualize its promise as an assistive technology for handicapped hearing individuals. This system can further be enhanced by including further classes on sign language and real-time gesture interpretation
Revolutionizing cancer classification: the snr-ogscc method for improved gene selection and clustering
Get PDFThis study presents the signal-to-noise ratio optimized gene selection and clustering for cancer classification (SNR-OGSCC) methodology, aimed at enhancing classification accuracy while reducing the dimensionality of gene expression data across various cancer types. Implemented on a standard computational setup, the SNR-OGSCC method combines advanced filtering, clustering, and machine learning techniques, demonstrating significant improvements in classification accuracy on seven cancer datasets: leukemia, colon cancer, prostate cancer, lung cancer, lymphoma, central nervous system (CNS) tumors, and ovarian cancer. Notably, our approach achieved perfect accuracies of 100% for leukemia, lung cancer, and ovarian cancer, with high accuracies of 98.4% for colon cancer, 99.1% for prostate cancer, 98.3% for lymphoma, and 99.7% for CNS tumors, while requiring as few as 4–5 genes for effective classification. These findings highlight the efficiency and robustness of the SNR-OGSCC methodology, suggesting its potential to identify meaningful biomarkers and improve personalized cancer treatment strategies. Further validation with larger datasets and biological experiments is essential to confirm its applicability in clinical settings
Hypovigilance detection based on analysis and binary classification of brain signals
Get PDFRoad safety has now become a priority for drivers and citizens alike, given its considerable impact on the economy and human life, which is reflected in the increase in the number of accidents worldwide. This increase is linked to a number of factors, drowsiness being one of the main causes that can lead to tragic consequences. Various systems have been developed to monitor the state of alertness. The main idea adopted in this paper is based on the integration of a biosensor to acquire the cerebral signal, then the processing and analysis of the characteristics required to detect the two states of the driver using intelligent machine learning algorithms. Two models were chosen to carry out this binary classification: The K-nearest neighbour (KNN) and logistic regression (LR) classifiers. The experimental simulation results show that the first model outperforms the second in terms of accuracy, with a percentage of 97.83% for k=3. This could lead to the development of a new safety machine brain system based on classification to control vehicle speed deceleration or activate self-driving mode in the event of hypovigilance
Intelligent cervical cancer detection: empowering healthcare with machine learning algorithms
Get PDFCervical cancer remains a significant global health issue, particularly in underdeveloped nations, where it contributes to high mortality rates. Early detection is critical for improving treatment outcomes and survival rates. This study employs machine learning (ML) algorithms to predict cervical cancer risk using a dataset from the University of California at Irvine (UCI), which includes demographic and clinical attributes such as age, sexual history, smoking habits, and medical history. After applying data preprocessing techniques, several classification algorithms, including logistic regression (LR), support vector machine (SVM), random forest (RF), decision tree, adaptive boosting (AdaBoost), and artificial neural networks (ANN), were trained and evaluated. The models were assessed using classification metrics such as precision, recall, and F1 score. Among the models, the ANN demonstrated the highest accuracy, achieving a score of 0.95. In addition, correlation analysis revealed significant relationships between various risk factors, providing insights into cervical cancer mechanisms and potential preventive measures. The study highlights the potential of ML in improving cervical cancer detection and patient outcomes, suggesting that advanced ML techniques can be valuable tools in healthcare research and clinical applications
Overcoming imbalanced rice seed germination classification: enhancing accuracy for effective seedling identification
Get PDFThis study aimed to automatically classify rice seedling germination on day seven using image analysis. The categories included normal, abnormal, and dead seeds. Due to the rarity of abnormal seedlings, capturing their images resulted in imbalanced data. To address this, abnormal categories were combined into a single class. We compared logistic regression, random forest, and deep learning models (VGG19, VGG16, Alex Net) for classification. Surprisingly, logistic regression achieved the highest accuracy (93.89%) and F1-scores (0.96 normal, 0.81 abnormal) despite the imbalanced data and complex task. The effectiveness of logistic regression for rice seedling classification with imbalanced data has been demonstrated in this novel research. Historically, deep learning models dominate image recognition, but our findings suggest simpler models can excel in specific scenarios, especially with limited data availability. This highlights the importance of selecting models based on data characteristics. The urgency for this research stems from the need for efficient and accurate rice seedling evaluation. Improved classification can enhance agricultural practices and optimize resource allocation
IoT-enabled Edge Impulse approach for heat stress prediction in outdoor settings
Get PDFSeveral international organizations of public health or countries have predicted the rise of heat-related illness cases due to climate change, which result high environment temperature. Previous studies of heat-related illness prediction using internet of things (IoT) and machine learning (ML) are mainly focusing on early detection or prediction of heat stroke incidence. To overcome the problem of heat stress prediction in outdoor settings, especially for an individual, the objective of this study is to identify a prediction method for heat stress using IoT technology and analyze the accuracy of the identified prediction model. Arduino nano 33 BLE sense with Bluetooth low energy (BLE) connectivity, HTS221 embedded environment temperature and humidity sensor, MLX90614 skin temperature sensor, and MAX30100 heart rate sensor were used to build IoT based wearable device. Besides, Python language is used for data pre-processing and data labelling after getting the sensor data from wearable device. Lastly, model training using neural network algorithms was directed in Edge Impulse. The result shows 94.6% of training accuracy with the loss of 0.27 and 90.22% of accuracy in testing set
Sign language recognition and classification using blended ensemble machine learning
Get PDFAn efficient sign language recognition system (SLR) is the most significant for hearing-impaired people for communication. The body movements and hand gestures are utilized to characterize the vocabulary in dynamic sign language. The SLR is a challenging problem because the computational model requires simultaneous spatial-temporal modelling for a number of sources. To overcome this problem, this research proposes the blended ensemble machine learning (ML) approaches for SLR. Initially, the Indian sign language (ISL) dataset is collected for evaluating the effectiveness of the model. Then, the pre-processing is done by using data augmentation and normalization techniques. Then, the pre-processed data is provided to the segmentation process which is done by using multi-threshold entropy function. Then, VGG-16 is used for the feature extraction process to extract the features and finally, classification is carried out using ensemble ML. An effectiveness of the proposed method is validated based on accuracy, precision, recall, and F1-score, wherein it achieves better results of 99.57%, 0.92%, 0.95%, and 0.99% as compared to the existing works like support vector machine (SVM) and convolutional neural network (CNN)