IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
A review of recent deep learning applications in wood surface defect identification
Get PDFWood is widely used in construction, art, and home applications due to its aesthetic appeal and favorable mechanical properties. However, environmental factors significantly affect the growth and preservation of wood, often leading to defects that can reduce its performance and ornamental value. Researchers have introduced machine vision and deep learning methods to address the challenges of high labor costs and inefficiencies in identifying wood defects. Deep learning has shown great success in image recognition tasks, yielding impressive results. This paper reviews previous work on deep-learning strategies for identifying wood surface defects. It also discusses data augmentation techniques to address limited defect data and explores transfer learning to enhance classification accuracy on small datasets. Finally, the paper examines the potential limitations of deep learning for defect identification and suggests future research directions
Application of the adaptive neuro-fuzzy inference system for prediction of the electrical energy production in Jakarta
Get PDFJakarta, as a rapidly growing urban area, faces challenges in balancing energy demand with supply while addressing environmental concerns associated with traditional energy sources. Electrical energy production prediction in urban environments like Jakarta is crucial for effective energy management, ensuring stable supply, and promoting sustainable development. The prediction of electrical energy production in Jakarta is critical for ensuring stable and sustainable energy supply. This research proposed the application of the adaptive neuro-fuzzy inference system (ANFIS) as a predictive tool specifically tailored for Jakarta's energy production prediction context. The research methodology used in this study is the ANFIS. Five levels make up the architecture of the ANFIS model: output, normalization, defuzzification, rule evaluation, and fuzzification. The fuzzification layer converts input variables into linguistic terms using membership functions, while the rule evaluation layer calculates the activation strength of each rule based on the input values. The predicted results of Jakarta electrical energy production from 2023 to 2028 are 65,288 GWh and there is an annual increase of 5.25%. The error contained in ANFIS is with a root mean square error (RMSE) value of 0.0001058% and a mean absolute percentage error (MAPE) value of 0.00875%
Semantic based medical visual question answering with explainable artificial intelligence
Get PDFThe medical visual question answering (MVQA) system takes the advantage of both computer vision (CV) and natural language processing (NLP) to accept the medical image and corresponding question as input and generates the respective answer as output. One step further, the MVQA system capable of generating the answer based on the semantics has a distinct place and hence semantic based medical visual question answering (SMVQA) system is proposed in this research. In SMVQA, the semantics for input image and question are generated using layerwise relevance propagation explainable artificial intelligence (LRP XAI) technique and the answer is derived using deductive reasoning method. For this, seven MVQA datasets are used for model creation, testing and validation. The training phase of the SMVQA system is implemented using VGGNet, long short-term memory (LSTM), LRP XAI, ResNet and bidirectional encoder representations from transformers (BERT) to generate a model file. Then the inference is derived in the testing phase based on the generated model file for the test set. Finally, the answer is derived from the inference using natural language toolkit (NLTK) library, term frequency-inverse document frequency (TF-IDF), cosine similarity, best match25 (BM25) techniques along with deductive reasoning. As a result, the proposed SMVQA system gives improved performance then the existing MVQA system especially for abnormality type samples
Automatic essay scoring: leveraging Jaccard coefficient and Cosine similarity with n-gram variation in vector space model approach
Get PDFAutomated essay scoring (AES) is a vital area of research aiming to provide efficient and accurate assessment tools for evaluating written content. This study investigates the effectiveness of two popular similarity metrics, Jaccard coefficient, and Cosine similarity, within the context of vector space models (VSM) employing unigram, bigram, and trigram representations. The data used in this research was obtained from the formative essay of the citizenship education subject in a junior high school. Each essay undergoes preprocessing to extract features using n-gram models, followed by vectorization to transform text data into numerical representations. Then, similarity scores are computed between essays using both Jaccard coefficient and Cosine similarity. The performance of the system is evaluated by analyzing the root mean square error (RMSE), which measures the difference between the scores given by human graders and those generated by the system. The result shows that the Cosine similarity outperformed the Jaccard coefficient. In terms of n-gram, unigrams have lower RMSE compared to bigrams and trigrams
Machine learning application for particle accelerator optimization-a review
Get PDFParticle accelerators receive significant attention from researchers. This machine consists of various interdependent elements, so it is complex. Efficient system tuning and diagnostics are essential for utilizing accelerator technology. In addition, machine learning (ML) has been applied in several applications. ML methods such as artificial neural networks, random forest, reinforcement learning, genetic algorithm, and Bayesian optimization have been used for accelerator optimization. The optimization of particle accelerators covers their performance and efficiency. This paper reviews the application of ML techniques in optimizing particle accelerators, highlighting their importance in addressing the complexity inherent in accelerator systems and advancing accelerator science and technology
Enhanced pre-broadcast video codec validation using hybrid CNN-LSTM with attention and autoencoder-based anomaly detection
Get PDFThis study presents a machine learning-based approach for proactive video codec error detection, ensuring uninterrupted television broadcasting for TV Laayoune, part of Morocco’s SNRT network. Building upon previous approaches, our method introduces autoencoders for improved anomaly detection and integrates data augmentation to enhance model resilience to rare codec configurations. By combining convolutional neural networks (CNNs) and long short-term memory (LSTM) networks with an attention mechanism, the system effectively captures spatial and temporal video features. This architecture emphasizes critical metadata attributes that influence video playback quality. Embedded within the broadcasting pipeline, the model enables real-time error detection and alerts, minimizing manual intervention and reducing transmission disruptions. Experimental results demonstrate a 97% accuracy in detecting codec errors, outperforming traditional machine learning models. This study highlights the transformative role of machine learning in broadcasting, enabling scalable deployment across diverse television networks
BioTapSync: revolutionizing data synchronization with human touch
Get PDFHuman Tap introduces a new way to ensure secure data transmission synchronization by integrating advanced technologies. Motivated by the need for secure and efficient communication, it uses both near field communication (NFC) and human field communication (HFC) to provide a wide range of secure communication solutions. The aim is to create a system with detailed specifications, including maximum coverage range, frequency of operation, type of communication, and data rate for each protocol. These specifications are tailored for various applications, such as credit card payments, e-ticket bookings, E-ZPass systems, and item tracking. A notable contribution of Human Tap is its ability to achieve microsecond-level accuracy for distances up to 2 cm by thoroughly analyzing the relationship between distance and time. This innovative synchronization method not only ensures a secure data transmission environment but also shows remarkable flexibility, effectively addressing the challenges of modern communication systems. Human Tap sets a new benchmark for secure and adaptable data transmission technologies, paving the way for future advancements in the field. The objective is to establish a robust and versatile data transmission method that can be adapted to a wide range of modern applications
Optimizing nonlinear autoregressive with exogenous inputs network architecture for agarwood oil quality assessment
Get PDFAgarwood oil is highly valued in perfumes, incense, and traditional medicine. However, the lack of standardized grading methods poses challenges for consistent quality assessment. This study proposes a data-driven classification approach using the nonlinear autoregressive with exogenous inputs (NARX) model, implemented in MATLAB R2020a with the Levenberg-Marquardt (LM) algorithm. The dataset, sourced from the Universiti Malaysia Pahang Al-Sultan Abdullah under the Bio Aromatic Research Centre of Excellence (BARCE) and Forest Research Institute Malaysia (FRIM), comprises chemical compound data used for model training and validation. To optimize model performance, the number of hidden neurons is systematically adjusted. Model evaluation uses performance metrics such as mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), coefficient of determination (R²), epochs, accuracy, and model validation. Results show that the NARX model effectively classifies agarwood oil into four quality grades which is high, medium-high, medium-low, and low. The best performance is achieved with three hidden neurons, offering a balance between accuracy and computational efficiency. This work demonstrates the potential of automated, standardized agarwood oil quality grading. Future research should explore alternative training algorithms and larger datasets to further enhance model robustness and generalizability
Modified zero-reference deep curve estimation for contrast quality enhancement in face recognition
Get PDFFace recognition systems remain challenged by variable lighting conditions. While zero-reference deep curve estimation (Zero-DCE) effectively enhances low-light images, it frequently induces overexposure in normal- and high-brightness scenarios. This study introduces modified Zero-DCE combined with three established enhancement techniques: contrast stretching (CS), contrast limited adaptive histogram equalization (CLAHE), and brightness preserving dynamic histogram equalization (BPDHE). Evaluations employed the extended Yale face database B and face recognition technology (FERET) datasets, with 10 representative samples assessed using the blind/referenceless image spatial quality evaluator (BRISQUE) metric. Modified Zero-DCE with BPDHE produced optimal enhancement quality, achieving a mean BRISQUE score of 16.018. On the extended Yale face database B, visual geometry group 16 (VGG16) integrated with modified Zero-DCE and CLAHE attained 83.65% recognition accuracy, representing a 6.08-percentage-point improvement over conventional Zero-DCE. For the 200-subject FERET subset, residual network 50 (ResNet50) with modified Zero-DCE and CLAHE achieved 67.41% accuracy. Notably, standard Zero-DCE with CLAHE demonstrated superior robustness in extremely low-light conditions, highlighting the illumination-dependent performance characteristics of these enhancement approaches
Data-driven clustering and prediction of high school graduation rates in Indonesia (2015-2023) using machine learning
Get PDFThis study aims to analyze the graduation rate of senior high school education in 34 Indonesian provinces during the period 2015-2023 and identify patterns of educational disparities between regions. To achieve the objectives, this study applies a neural network to predict education completion patterns based on historical data, then the prediction results are analyzed using K-means clustering technique utilizing the elbow method to select the ideal number of clusters. The clustering results show three categories of provinces based on education completion rates: high, medium, and low. The provinces with high completion rates, generally, supported with good education infrastructure and effective policies, while the medium category faces challenges in resource distribution, but still potentially improve. In contrast, the low category suffers from limited access, geographical constraints, and socio-economic disparities. This research contributes to education policy-making by offering a machine learning-based approach to understanding education disparities between regions. The new insight offered by this study lies in the integration of neural network and K-means clustering in mapping education completion rates to support strategies for improving access and quality of education in Indonesia