EAI Endorsed Transactions on Intelligent Systems and Machine Learning Applications

Enhancing Object Recognition Through a Novel Adaptive Recognition Technique (ART) Framework

Hewa Majeed Zangana — 2025-10-07

Object recognition is a critical capability in various computer vision applications, but traditional approaches often struggle with complex, real-world scenarios. This paper introduces a novel Adaptive Recognition Technique (ART) framework to enhance object recognition performance. The proposed ART framework leverages adaptive learning mechanisms to more accurately identify objects, even in the presence of variations in size, orientation, and environmental conditions. Through a series of experiments on benchmark datasets, the ART framework demonstrated significant improvements in recognition accuracy compared to existing methods. Key innovations include the integration of unsupervised feature learning, dynamic model adaptation, and ensemble-based decision making. The results suggest that the ART framework offers a promising approach to advancing the state-of-the-art in object recognition, with potential applications in areas such as autonomous vehicles, surveillance, and image analysis. Further research is underway to expand the capabilities of the ART framework.

Timing for securing the biometric template transformation based on supervised learning using Double Random Phase Encoding Method

Mahmoud Nasr — 2025-06-18

Background: Among optical encryption techniques, Double Random Phase Encoding (DRPE) is one of the most widely used. Individual identities and the process of recognition remain essential to ensuring proper data access security.
Aim: The study aims to optimize an approach that ensures the significant performance effectiveness of the cancelable biometric methods for different templates and the associated time taken to transform biometric data.
Problem: This study is majorly concerned about the performance effectiveness of cancelable biometric methods that measure the likelihood that an authorized effort may be mistakenly rejected as unauthorized. Also, when compromised, several non-renewability safety challenges arise, and insufficient matching performance templates are required to build a security protection method.
Method and material. The study uses supervised learning for the Double Random Phase Encoding Method (DRPE), a 4F optical encryption system, and 20 randomly chosen photos from the ORL database of faces.
Results. The result based on the supervised learning for the Double Random Phase Encoding Method revealed false positive rates for both the fingerprint and face templates.
Conclusion. The study concluded that the performance effectiveness of the cancelable biometric in this study has a false positive rate likelihood that an authorized effort may not be mistakenly rejected as an unauthorized one.

A Machine Learning-based approach to predicting tuberculosis in the Democratic Republic of Congo

Pierre Tshibanda wa Tshibanda — 2025-07-16

INTRODUCTION: Tuberculosis remains a public health problem in Democratic Republic of Congo (DRC), despite advances in Machine Learning for the prediction of this disease. However, existing models are often adapted to Asian contexts and do not take into account the specific epidemiological and social characteristics of the DRC. Given this shortcoming, our study explores a Machine Learning approach specifically designed to improve the prediction of tuberculosis in the Congolese population.
OBJECTIVES: Our problem is based on the following question: "What approach, based on Machine Learning and specific to the population of DRC, is likely to improve the prediction of tuberculosis?" To answer this, we adopted an exploratory paradigm with a sequential mixed design (qualitative and quantitative). The study was conducted on a sample of 1505 patients and six healthcare professionals in the health zones of Lubumbashi and Nzanza.
METHODS: The data was collected using questionnaires and semi-structured interviews, then analysed using bivariate and multivariate approaches.
RESULTS: The results show that incorporating Congolese specificities into Machine Learning models significantly improves the prediction of tuberculosis. Of the models tested, Random Forest and Decision Tree performed best in terms of precision, recall, F1-score and AUC, while Voting Classifier, Stacking and Adaboost showed a good compromise between precision and robustness.
CONCLUSION: This study highlights the need to develop predictive models adapted to the local context in order to improve tuberculosis control in DRC. We propose an optimised model incorporating characteristics specific to the Congolese population, with a possible large-scale application to improve detection and prevention of the disease.

Al-Driven Qualitative Research in Smart Cities: Enhancing Emotional Resilience in Youth and Children

Edwin Gerardo Acuña Acuña — 2025-09-29

INTRODUCTION: Industry 5.0 has brought advanced AI-driven technologies into qualitative research and data analysis, particularly in systems that are very important to the purpose. This research examines the use of AI algorithms to evaluate emotional resilience in kids and children in smart cities. The study underscores Al's role in qualitative research to substantiate the efficacy of these algorithms in assessing emotional resilience and advocating for interventions that improve emotional well-being. The main goal of this research is to see how accurate and reliable AI algorithms are when they measure emotional resilience. The goal of the project is to leverage these technologies to make treatments that make kids and teens in smart cities feel better emotionally, which will help them grow up in a caring environment.

METHODOLOGY: A quantitative, descriptive, and exploratory methodology is used, using data gathered from children to examine emotional reactions via deep neural network models. These models are designed to find levels of resilience with amazing accuracy, sensitivity, and specificity, with the goal of getting accuracy rates above eighty percent.

RESULTS: The results indicate that AI-driven technology may provide comprehensive qualitative insights into the emotional resilience of adolescents and children. The research underscores the capacity of these technologies to provide personalized treatments and assistance, hence improving emotional well-being in smart city contexts. The findings indicate that AI might enhance emotional resilience, facilitate early detection of emotional problems, and enable prompt assistance. The suggested model was able to find emotional resilience with 94% accuracy, 92% sensitivity, 88% specificity, and 95% AUC. These results demonstrate the efficacy of AI-driven approaches in the early detection of emotional problems among adolescents and teenagers inside smart city environments. The research shows that AI technologies are very important for figuring out how to help kids and teens become more emotionally strong. It backs the employment of these technologies in the public health and education systems of smart cities to help kids develop emotionally. This plan makes it simpler to get in early and helps create a strong, supportive community.

Hybrid Template Matching and Faster R-CNN for Robust Multimodal Object Detection

Hewa Majeed Zangana — 2025-10-01

This paper introduces a hybrid object detection framework that integrates template matching with the Faster R-CNN deep learning algorithm to improve robustness in challenging conditions such as occlusion, clutter, and low resolution. The novelty of this work lies in systematically combining a traditional template-matching branch with a two-stage detector, enabling the system to capture predefined structural cues alongside learned deep features. The proposed score-based fusion mechanism further refines detections by weighting outputs from both branches. Experimental results on COCO and LASIESTA datasets show that the hybrid model achieves an F1 score of 88.6% and a mAP@0.75 of 69.4%, surpassing both template-only and Faster R-CNN-only baselines. These findings highlight the effectiveness of the hybrid strategy in enhancing detection accuracy and robustness while maintaining practical computational efficiency.