Conference papers
Permanent URI for this communityhttps://rda.sliit.lk/handle/123456789/4889
Browse
16 results
Search Results
Item Embargo Hybrid Deep Learning Approach Using YOLO and Spatial Transformers for Kinesthetic Math Education in Low-Resource Settings(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Sellapperuma M.S; Wasana, K.H. I; Dhananjana, B.K. T; Anuruddhika, D. L. N; Krishara, J; Wijendra, DThis research presents a hybrid deep-learning system to enhance kinesthetic math education for Grade 4 and 5 students in low-resource Sri Lankan classrooms, addressing addition, subtraction, and time-telling skills in Sinhala. Integrating fine-tuned You Only Look Once version 11 (YOLOv11) for real-time abacus recognition and pre-trained Spatial Transformer Network (STN) with Residual Network 50 (ResNet50) for analog clock analysis, the system bridges physical manipulatives with digital feedback. Optimized for low-cost hardware via post-training quantization, it reduces model sizes by up to 56.6% and inference times by 45.9%, enabling deployment on standard PCs. The methodology employs YOLOv11 to detect abacus beads (mean Average Precision, mAP50-95: 0.898) and STN + ResNet50 to correct clock perspectives (95.2% accuracy on SynClock), delivering immediate Sinhala feedback through a culturally adapted interface. The system was evaluated on an Intel Core i5 PC with 8GB RAM and achieved sub-second inference (YOLOv11: 170.1 ms, STN + ResNet50: 32 ms post-quantization) while retaining accuracy. A pilot study with 20 students showed a 25% improvement in arithmetic scores and 30% in time-telling accuracy, with 85% reporting higher engagement. These findings demonstrate the system's efficacy in boosting numeracy and interaction in resource-scarce settings, aligning with the Visual, Aural, Read/Write, Kinesthetic (VARK) framework. Despite challenges like hardware constraints and real-world variability, this scalable, offline-capable solution offers a novel approach to educational technology, with potential for broader deployment in Sri Lanka and similar contexts, addressing gaps in artificial intelligence-driven kinesthetic learning tools. ©Item Embargo FITGEN - AI Smart Fitness Companion: An Integrated Multi-Modal Approach to Personalized Health and Fitness Management(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Gunawardena S.N.A; Jayamali P.L.W; Methsandi K.A.N.; Gamage R.P.D.D; Wijendra, D; Krishara, JThis research presents FITGEN, a comprehensive AI-powered fitness companion system that integrates four core modules: AI Gym Trainer with Gamification, Medical Guidance System, AI Nutritionist, and Social Bridge for inclusive fitness engagement. Unlike existing solutions that address singular aspects of fitness technology, FITGEN provides a holistic approach combining real-time pose detection, chronic disease management, personalized nutrition planning, and accessibility features for special abled youth. The system utilizes Flutter framework with Firebase backend, implementing Google ML Kit for pose estimation, TensorFlow models for nutrition recommendations, and ESP32-based IoT sensors for vital sign monitoring. Comprehensive evaluation demonstrates 93.5% pose detection accuracy, 96% nutrition recommendation accuracy, 73% user retention after 4 weeks, and significant improvements in accessibility compliance. The integrated approach addresses critical gaps in existing fitness technologies while maintaining high user engagement and safety standards for diverse user populations including those with chronic conditions and developmental disabilities.Item Embargo Context-Aware Behavior-Driven Pipeline Generation(Institute of Electrical and Electronics Engineers Inc., 2025-04-24) Gunathilaka, P; Senadheera, D; Perara, S; Gunawardana, C; Thelijjagoda, S; Krishara, JAn efficient CI/CD process is crucial for modern software teams, but manual pipeline creation is error-prone and requires high DevOps expertise, slowing deployment speed and reducing productivity. This research introduces a context-aware, behavior-driven approach to fully automating CI/CD pipeline generation by analyzing GitHub user activity patterns. The proposed solution utilizes a historical analysis of repository events, developer contributions, and workload distribution to dynamically generate pipelines and assign reviewers to pull requests based on expertise. Unlike previous template-based and generative AI solutions that require manual intervention, our approach leverages pattern recognition and adaptive decision-making to continuously refine automation. This paper presents the methodology behind data collection, analysis, and pipeline generation, demonstrating its effectiveness in reducing human effort while improving software delivery efficiency. This research highlights how behavior-driven automation streamlines the complexity of CI/CD pipeline creation, enabling more adaptive and intelligent systems that effectively respond to the evolving needs of software development teams.Item Embargo Context-Aware Incremental Learning Pipeline Triggers(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Perara, S; Senadheera, D; Gunathilake, P; Gunawardana, C; Thelijjagoda, S; Krishara, JModern Continuous Integration/Continuous Delivery pipelines require intelligent and adaptive learning mechanisms to enhance software deployment stability and efficiency. While previous studies have introduced Pull Request prioritization techniques and predictive models for build failures, they do not offer real-time recommendations based on Pull Request criticality and build stability. This research presents a GitHub app based on reinforcement learning for intelligent Pull Request prioritization and deployment advisory in Continuous Integration/Continuous Delivery workflows. T he system combines natural language processing with contextual features to dynamically determine whether a Pull Request should be prioritized for early deployment or flagged for urgent review based on its priority level and associated build failure risk. Due to limited labeled feedback data, the study employs a hybrid approach using supervised learning for initial classification, enhanced by reinforcement learning for continuous adaptation. Using a dataset of 682148 Pull Requests from 28835 repositories, experimental results demonstrate significant improvements in deployment efficiency, achieving 92.01% accuracy in Pull Request classification with strong developer alignment. This research presents a comprehensive solution for optimizing Continuous Integration/Continuous Delivery pipelines through intelligent Pull Request prioritization with deployment advisory.Item Embargo DiverseMind: An Integrated Framework for Children with Multi-Dimensional Challenges as Slow Learners(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Jayasundara, H; Neewin, S; Kiriwaththuduwa, C; Herath, R; Krishara, J; Thelijjagoda, SEducation systems worldwide struggle to support slow learners, who face difficulties in traditional classrooms due to learning challenges in writing, mathematics, attention, and memory. Slow learners, characterized by an Intelligence Quotient (IQ) between 70 and 85, require additional time and adaptive learning methods to grasp concepts effectively. However, existing educational frameworks lack comprehensive screening and targeted interventions. This research introduces "DiverseMind", an integrated framework designed to identify and assist slow learners among Grade 4 primary school children in Sri Lanka using advanced Machine Learning (ML) algorithms, image processing, and multi-model architecture. The system evaluates academic abilities through four key assessments of writing skills, mathematical proficiency, attention span, and short-term memory. A Convolutional Neural Network (CNN) based model, achieving a training accuracy of 98% combined with a Python-based weighted condition function, classifies writing accuracy, while Decision Tree (DT) classifiers analyze mathematical capabilities with 98% accuracy. Attention span is assessed using facial landmark detection, gaze tracking, and emotion recognition, where the CNN model trained on 28,709 images achieved a training accuracy of 80%. Short-term memory is evaluated through ML driven cognitive tasks, with the DT model achieving 99% accuracy. In addition to comprehensive assessments and interventions, the system provides a dedicated dashboard for the teachers to monitor the student progress. By integrating gamification and AI-driven learning analytics, "DiverseMind"promotes inclusive education and bridges the gap in support for slow learners, ensuring they receive the necessary resources to reach their full potential.Item Embargo CINNOVA: Advancing Sustainable Cinnamon Farming through AI and Collaborative Solutions(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Bandara D.; Fernando M.K.K.L; Senadeera N.A.D.N; De Silva R.C.T; Wijendra, D; Krishara, JCinnamon is one of the most economically significant export crops in Sri Lanka. However, its cultivation is challenged by plant diseases, nutrient deficiencies, and inefficient harvesting practices, which reduce yield quality and productivity. Traditional methods for identifying plant health issues are time-consuming and require expert evaluation, often inaccessible to rural farmers. To address these limitations, this study introduces an AI-driven intelligent monitoring system for sustainable cinnamon cultivation, a mobile-based solution specifically designed to enhance cinnamon farming practices. The system leverages Artificial Intelligence (AI), Deep Learning (DL), and Image Processing techniques to support real-time plant health diagnostics. It integrates multiple AI-powered components for early detection of bark diseases such as Rough Bark Disease (RBD) and Canker Disease (CD) using a contrastive learning-based model, severity prediction of leaf diseases, including Leaf Gall and Leaf Blight, using the YOLO model, identification of nutrient deficiencies, particularly Magnesium and Potassium, through transfer learning and prediction of cinnamon bark maturity and quality grades utilizing spatial attention mechanisms based on diameter and color. Each model is optimized for mobile deployment to provide real-time feedback and enable efficient decision-making. This AI-driven approach enhances disease management and improves yield quality while promoting sustainable, data-driven cinnamon cultivation in Sri Lanka.Item Embargo Grade 6 Coding Education through Interactive Learning in Sri Lanka(Institute of Electrical and Electronics Engineers, 2025-12-08) Nallaperuma, S; Perera, A; Nanayakkara, N; Hansaja, D; Krishara, J; Wijendra, DThis study proposes an AI-powered Adaptive Learning Management System (LMS) to enhance coding education for school students. The system integrates six AI/ML models and FaceNet-based facial authentication to deliver adaptive learning pathways, real-time feedback, and personalized assessments. It begins with an initial benchmarking assessment across seven categories, after which the Exam Difficulty and Question Count Model and a fine-tuned GPT-2 generator collaboratively create customized multiple-choice questions based on each learner's capability. Learning is supported through skill prediction, dynamic difficulty monitoring, and emotional state detection to ensure both cognitive and emotional engagement. Students scoring above 60% advance directly to coding curricula, while others receive personalized scaffolding through multimodal learning-style-based modules. Teachers benefit from AI-powered analytics tools that offer continuous feedback, weak-topic detection, and pedagogical guidance to enhance teaching effectiveness. By combining authentication, adaptive assessment, emotional intelligence, and teacher support, this study presents a comprehensive framework to democratize coding education and foster resilience, motivation, and mastery among students aged 10-12.Item Open Access Intelligent Systems for Comprehensive Dog Management(Association for Computing Machinery, 2025-06-28) Katipearachchi, M.E; Sachethana, O; Gunawardena, G.N.A; Ruwanara, D.C; Krishara, J; Kasthurirathna, DIn recent years, the integration of advanced technologies with canine welfare has gained significant attention, leading to the development of comprehensive platforms for dog management. The "Research Pooch-Paw"initiative addresses the multifaceted needs of dog owners and stray dog populations through an innovative platform that incorporates machine learning, wearable sensors, and real-time data processing. The platform facilitates early disease detection, behaviour analysis, and health monitoring using IoT-enabled devices, and provides personalized care guidance. Additionally, it includes features for stray dog identification and emergency response using deep learning algorithms and image processing techniques. The research underscores the potential of leveraging modern technology to enhance the quality of life for dogs and improve the effectiveness of canine welfare strategies.Item Embargo Knowledge Graph-Based AI Framework for Predicting Nutritional and Health Impacts of Food Ingredients(Institute of Electrical and Electronics Engineers Inc., 2026-08-04) Dakshina P.D.S.D; Rupasighe W.A.R.K; Waduge N.P; Nimsitha M.V.T; Tissera, W; Rathnayake, S; Krishara, JThe increasing complexity of modern food products and dietary supplements has made it challenging for both consumers and healthcare professionals to interpret nutritional information and assess the potential health risks associated with these products. Modern food labeling schemes provide static and fragmented information and cannot effectively capture the relationships between different ingredients, nutrients and their health effects. In this study, a new AI-based framework named Food Health Risk Analyzer has been proposed that utilizes KGs, GNNs, RAG and a dose-response module based on consumption quantities to perform the dynamic, explainable and evidence-based prediction of food-related health risks. The model uses heterogeneous data in order to analyze the relationships between ingredients and diseases to predict potential health risks while generating scientifically supported explanations as well. The experimental evaluation has shown high prediction accuracy with a micro-F1 score of 0.88 and AUC of 0.85 which shows that the framework surpasses conventional machine learning baseline models. In addition to that, the use of RAG has helped in improving the interpretability of predictions through evidence-based natural language explanations whereas dose-response module improves the practical relevance of risk assessment by considering the consumption quantities of ingredients.Item Embargo Throat AI - An Intelligent System For Detecting Foreign Objects In Lateral Neck X-Ray Images(Institute of Electrical and Electronics Engineers Inc., 2025) Baddewithana, P; Krishara, J; Yapa, KForeign Object ingestion is a commonly encountered medical condition within the Ear, Nose, and Throat clinical domain. Timely and accurate detection of such objects is vital, as it often guides the need for surgical intervention. Among the available imaging techniques, lateral neck X-rays are the most widely used radiographs to visualize and assess the presence of FOs in the throat. However, manual interpretation of these images can be time-consuming and subject to human error, potentially leading to misdiagnosis or delayed treatment. This research presents a deep learning-based software solution, deployable via web and mobile platforms, aimed at assisting medical professionals with the automated detection of FOs in lateral neck X-rays. The system leverages state-of-the-art YOLO object detection models, specifically evaluating novel versions such as YOLO-NAS-s, YOLOv11s, and YOLOv8s-OBB to ensure high detection accuracy and deployment efficiency. The best-performing model, YOLO-NAS-s, achieved a validation accuracy of 96.3%. For deployment, the model was hosted on the Roboflow platform and accessed via a FastAPI-based middleware server. Performance evaluation showed an average inference time of approximately 2 seconds and a memory footprint of around 100 MB on standard computing hardware, demonstrating its suitability for integration into resource-constrained clinical environments. This setup highlights the system's lightweight design and real-world applicability. Training, evaluation, and testing of the deep learning models were conducted using a dataset curated from public local healthcare institutions and online medical imaging repositories.