IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Implementation and evaluation of Heskes self organizing map counter propagation network for face recognition
Get PDFFace recognition has attracted a lot of interest in the fields of computer vision and pattern recognition given its extensive applications in security, surveillance, and human-computer interaction. Many linear and non-linear classifiers have been introduced to bring about effectiveness in face recognition, however, the problem of occlusion, light conditions and changes in face persist. The Heskes-self-organizing map (SOM) counter propagation network (CPN) model leverages the competitive learning and self-organizing features of SOM CPN with Heskes layer to improve the effectiveness and accuracy of face recognition systems. Heskes-SOM CPN was implemented and evaluated on MATLAB R2016a using 600 images captured with the aim of digital camera. The implemented model was trained with 360 face images and tested with 240 face images using accuracy, sensitivity, specificity, and false positive rate as performance metrics at four distinct threshold values of 0.23, 0.35, 0.50, and 0.75. The major objective of the research was achieved by investigating with 50×50 and 200×200 face dimensions. Empirical results and statistical evidence established that Heskes-SOM CPN has high accuracy of approximately 97.92%, high specificity of 98.33%, high sensitivity of 99.44%, and a very low filter performance rating (FPR) of 1.67%. Therefore, Heskes-SOM CPN is presented as a novel CPN model for face recognition
A comparative analysis of optical character recognition models for extracting and classifying texts in natural scenes
Get PDFThis research introduces prior-guided dynamic tunable network (PDTNet), an efficient model designed to improve the detection and recognition of text in complex environments. PDTNet’s architecture combines advanced preprocessing techniques and deep learning methods to enhance accuracy and reliability. The study comprehensively evaluates various optical character recognition (OCR) models, demonstrating PDTNet’s superior performance in terms of adaptability, accuracy, and reliability across different environmental conditions. The results emphasize the need for a context-aware approach in selecting OCR models for specific applications. This research advocates for the development of hybrid OCR systems that leverages multiple models, aiming to arrive at a higher accuracy and adaptability in practical applications. With a precision of 85%, the proposed model showed an improved performance of 1.7% over existing state of the arts model. These findings contribute valuable insights into addressing the technical challenges of text extraction and optimizing OCR model selection for real-world scenarios
The crucial role of artificial intelligence in addressing climate change
Get PDFAddressing climate change is one of the fundamental priorities at a global level, given its significant impact on both the environment and society. This systematic literature review explores the role of artificial intelligence (AI) in addressing climate change. It identified applications, contributions to predicting extreme events, techniques used, ethical challenges, and associated biases. The rapid systematic literature review (RSL) was conducted using databases such as Scopus, Dimensions, directory of open access journals (DOAJ), and IEEE Xplore. The preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement was used to ensure the completeness and transparency of the analysis. 40 articles were selected that were published between 2018 and 2023 and addressed AI in climate change. The findings show that AI is being used to predict and mitigate extreme climate events, estimate the greenhouse effect, and predict temperatures. In addition, innovative techniques such as hybrid machine learning models, convolutional neural networks, artificial neural networks, support vector machines, and logistic regression. In conclusion, AI offers a promising approach to addressing climate change, with transformative potential in predicting and mitigating its effects. However, continuous ethical considerations are required to guarantee its conscientious and efficient utilization
Deep learning-based classifier for geometric dimensioning and tolerancing symbols
Get PDFThis research investigates the recognition of geometric dimensioning and tolerancing (GD&T) symbols using a deep learning model for object detection. GD&T, playing a pivotal role in engineering and manufacturing, provides essential specifications for product design and production. Manual processes for GD&T are often time-consuming and error prone. The study demonstrates outstanding accuracy in automating GD&T symbol recognition in engineering applications using YOLOv8. A carefully curated dataset, encompassing a wide range of GD&T symbols, was employed for training and evaluating the model. The YOLOv8 architecture, renowned for its robust performance, was meticulously fine-tuned to cater to the specific requirements of GD&T symbol detection. This research not only addresses the challenges in manual GD&T processes but also showcases practical implications for improved quality control and streamlined engineering workflows. By automating GD&T symbol recognition, this study contributes to the efficiency and precision crucial in the engineering and manufacturing domains
Efficient reconfigurable parallel switching for low-density parity-check encoding and decoding
Get PDFIn the evolution of next-generation communication systems, the demand for higher data integrity and transmission efficiency has brought low-density parity-check (LDPC) codes into focus, particularly for their error-correcting prowess. Traditional LDPC encoding and decoding techniques, such as the belief propagation (BP), Min-Sum, and Sum-Product algorithms, are hampered by high computational complexity and latency. Our research introduces a groundbreaking approach: an efficient, reconfigurable highspeed parallel switching operation for a complexity-optimized low-density parity-check encoding and decoding model (CoLDPC-EC). This method leverages advanced parallel processing and reconfigurable computing to drastically enhance operational speed and efficiency. It significantly outperforms conventional algorithms by optimizing key parameters like decoding throughput and power consumption, ensuring swift, energy-efficient error correction ideal for cutting-edge communication technologies. Our comparison with traditional methods underscores our solution's superior speed, flexibility, and efficiency, promising a leap forward in reliable, highspeed data transmission for next-generation networks. As per the simulation analysis, the proposed system shows better performance compared to conventional methods by 10.35%, 3.56%, and 2.36% in terms of decoding throughput, power consumption, and energy efficiency error correction, respectively
Accurate prediction of chronic diseases using deep learning algorithms
Get PDFIn this paper, the researchers studied the effects of different activation functions in hidden layers and how they impact the overfitting or underfitting of the model in the multiclass prediction of chronic diseases. This paper also evaluated the effects of varying the number of layers, and hyperparameters and its impact on the accuracy of the model and its generalization capabilities. It was found that exponential linear unit (ELU) does not have a significant advantage over rectified linear unit (ReLU) when used as an activation function in the hidden layer. Additionally, the performance of softmax function, when used in the output layer, is the same as a classic sigmoid output activation function. In terms of the ability of the model to generalize, the researchers achieved a classification accuracy of 100% when the trained model was used to predict unseen data. Through this research, the researchers should be able to assist medical professionals and practitioners in Oman in the validation and diagnosis of chronic diseases in clinics and hospitals
A novel model to detect and categorize objects from images by using a hybrid machine learning model
Get PDFAs humans, we can easily recognize and distinguish different features of objects in images due to our brain’s ability to unconsciously learn from a set of images. The objectives of this effort are to develop a model that is capable of identifying and categorizing objects that are present within images. We imported the dataset from Keras and loaded it using data loaders to achieve this. We then utilized various deep learning algorithms, such as visual geometry group (VGG)-16 and a simple net-random forest hybrid model, to classify the objects. After classification, the accuracy obtained by VGG16 and the hybrid model was 84.7% and 89.6%, respectively. Therefore, the proposed model successfully detects objects in images using a simple net as a feature extractor and a random forest for object classification, achieving better accuracy than VGG16
Diabetes mellitus diagnosis method based random forest with bat algorithm
Get PDFDiabetes mellitus (DM) is a very dangerous disease and can cause various problems. Early diagnosis of DM is essential to avoid severe effects and complications. An affordable DM diagnosis method can be developed by applying machine learning. Random forest (RF) is a machine learning technique that is applied to develop a DM diagnosis method. However, the optimization of RF hyperparameters determines the performance of RF approach. Swarm intelligence (SI) could be used to solve the hyperparameter optimization problem on RF. It is robust and simple to be applied and doesn’t require derivatives. Bat algorithm (BA) is one of SI techniques that gives a balance between exploration and exploitation to find a global optimal solution. This article proposes developing an RF-BA-based technique for diagnosing DM. The results of the experiment demonstrate that RF-BA can diagnose DM more accurately than conventional RF. RF-BA has higher performance compared to RF-particle swarm optimization (PSO) in terms of computational time. The RF-BA also are able to solve the overfitting problem in the conventional RF. In the future, the proposed method has a high chance of being implemented for helping people with early DM diagnosis with high accuracy, low cost, and high-speed process
Two-step convolutional neural network classification of plant disease
Get PDFIndonesia is primarily an agricultural country, with farming being the primary source of income for most of its people. Unfortunately, crop production is vulnerable to plant diseases, which are usually caused by plant pests, resulting in a reduction in both the quantity and quality of the expected harvest. In addition to the large number of classes to predict, detecting and accurately classifying each disease on different plants can be difficult. We believe that limiting the number of classes to identify may improve classification accuracy. Thus, in this research, we propose a new approach, two-step convolutional neural network (CNN), which reduces the number of classes with a two-step classification approach. To begin, we identify the number of classes that can be reduced by categorizing them into different characteristics, namely, plant type classification and plant condition classification. Second, we deal with unbalanced datasets, which can result in poor performance, if overlooked. Finally, we compare the proposed two-step CNN to baseline CNN in terms of efficiency and effectiveness. Extensive experiments show that the two-step CNN outperforms the baselines, CNN and jellyfish-residual network (JF-ResNet), increasing accuracy by 4% and 2% to 99%, respectively. In addition, we also provide a simulation evaluation to ensure that this approach is applicable
Detection of location-specific intra-cranial brain tumors
Get PDFMutations or abnormalities in genes can occasionally cause cells to grow uncontrolled, resulting in a tumor, which is very dangerous. These are the most prevalent cancer causes. They are caused by significant damage to genes in a specific cell during a person's existence. Brain tumors are increasing rapidly, majorly brain tumor cases in the US are projected to rise from 27,000 in 2020 to 31,000 in 2023 at an annual growth rate of 1.5%, all the cases are rising because of the detection of the tumors in the late phase. Thus, it needs the hour to create something which can solve this anomaly and help us detect the tumor rapidly and efficiently. While major research papers on brain tumor detection mainly focus on the detection and classification of the tumors, the presented research aims to first detect the tumor using pre-recognized photos using machine learning object detection models. Then after successful detection of the tumor, the study team plans to determine its precise coordinates and display the tumor and its location in the picture