Recent Submissions
Contextual Bias Score - A Quantitative Metric for Gender Bias Detection in Visual Datasets
(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Ekanayake, A; Mudalige, T; Dhanawardhana, I; Jayawardena, S; Haddela, P. S; Shyamalee, T
Artificial Intelligence (AI) systems in computer vision often inherit contextual biases from training datasets, resulting in unintended gender associations during visual recognition. This paper introduces the Contextual Bias Score (CBS), a novel metric for quantifying gender bias in image datasets by evaluating object prominence, 3D spatial relationships, and scene-level semantics. Unlike conventional approaches that rely on co-occurrence statistics, CBS integrates object-person spatial features, relative size, depth, and distance with CLIP-based scene embeddings to assess how visual context influences bias. The framework is evaluated across nine experimental approaches using eight COCO-derived datasets, leveraging SHAP-based interpretability, PCA-based feature weighting, and Ridge regression modeling. Approach 9 (SHAP + Ridge) achieves 85% agreement with human-annotated misclassification labels and strong score stability. Validation includes statistical significance testing, human perception alignment, and cross-approach correlation analysis. CBS offers a dataset-independent, interpretable tool for auditing contextual bias in high-impact domains. Limitations include binary gender assumptions and computational overhead (~90 seconds per image), with future work aimed at extending the framework to non-binary and culturally diverse contexts.
PredictiveOps : Advanced Predictive Maintenance System for Industrial Machinery
(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Yapa, I; Gunawardana, J.V; Madhuwantha, H; Gunasekara, S; De Silva, H
Predictive maintenance is transforming industrial equipment management through the application of machine learning (ML) and the Internet of Things (IoT) for pre-failure detection. This paper investigates the PredictiveOps system that minimizes downtime, lowers maintenance expenditures, and increases operational effectiveness. This research puts the IoT sensors' function for the monitoring of such key parameters as vibration, temperature and rotational speed into the limelight, making condition-based maintenance possible. Anomaly detection with explainable AI ensures actionable output for maintenance staff. Predictive models also ensure minimal spare parts inventory, preventing undue stockholding. Findings indicate that predictive maintenance improves equipment reliability, prolongs asset life, and minimizes industrial waste. Challenges such as sensor noise and data imbalance still exist, calling for additional research. Through the transition from reactive to proactive measures, PredictiveOps is advancing Industry 4.0, promoting efficiency and sustainability.
Detecting and Monitoring Agricultural Lands in Sri Lanka Using Remote-Sensing and Deep Learning
(Institute of Electrical and Electronics Engineers Inc., 2025-12-17) Fernando P.S.R; Saldin M.Z; Perera B.W.L; Premarathna M.P.H.M; Abeywardhana, L; Dassanayake, T
Accurate agricultural assessment is essential for improving productivity and ensuring food security in regions like Sri Lanka. This study presents an integrated framework using remote sensing and deep learning to enhance precision agriculture. Multispectral satellite imagery from Sentinel-2 and Landsat, processed via Google Earth Engine (GEE), supports a soft-label Convolutional Neural Network (CNN) for identifying crop types and phenological stages. Geospatial techniques enable precise area mapping and vegetation health evaluation, while Long Short-Term Memory (LSTM) networks model temporal growth dynamics for yield forecasting. The system's web interface provides real-time analytics, interactive maps, and automated alerts on crop performance and underutilized lands, enabling timely, data-driven decisions. This approach offers a scalable, technology-driven pathway for sustainable land management and improved agricultural resilience in Sri Lanka.
Early Autism Screening via Multimodal Flat Affect Recognition in Voice and Facial Cues
(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Siriwardana K.D.S.P; Prasadi P.G.K.R; Vidyaratna K.R.K; Dulangi G.K.N; Nawarathne, M; Abeygunawardhana, P. K. W
Flat affect, characterized by reduced facial and vocal emotional expressiveness, is one of the key indicators of Autism Spectrum Disorder (ASD) in children. The present article suggests a novel automatic system for the early identification of flat affect in 8-11 years old children using a multimodal fusion of facial movement, head pose, global attention, and voice-based emotion features such as pitch, volume, and speech patterns. A decision model integrates all these features to enhance the detection of emotional expressiveness deficits. Compared to previous research on a single modality, this multimodal approach increases detection reliability. Experimental evaluation demonstrates high performance with 90% accuracy, 92.86% precision, and 86.67% recall. The proposed system provides an effective, objective tool to assist in early ASD symptom detection and intervention for psychologists, educators, and parents.
Bedtime Smartphone Use and Sleep: Exploring Scalable Solutions for Long-Term Well-Being
(Institute of Electrical and Electronics Engineers Inc., 2025-07-03) Jayamanne J.A.P.M; Rajapaksha R.A.N.M; Anjalee P.A.D.L; Siyambalapitiya H.N; Wijendra, D. R
Smartphone usage is widespread among young adults, yet its excessive use at bedtime raises concerns about its impact on sleep quality. This study investigates the effects of bedtime smartphone use and proposes an AI-driven system to mitigate its negative influence. The system integrates machine learning models and personalized intervention techniques to enhance sleep hygiene. Data collection involved young adults (20-30 years old) analyzing sleep patterns, common disruptors, and user preferences for interaction, including camera-based emotion detection and chatbot-driven guidance. The proposed model leverages multi-task learning for sleep pattern prediction, real-time app usage monitoring, and emotion recognition to provide adaptive recommendations. Additionally, gamification strategies, such as progress tracking and reward-based engagement, encourage adherence to healthier sleep routines. The findings demonstrate the potential of technology-driven interventions in addressing digital-age sleep challenges. By integrating AI-powered insights, personalized strategies, and interactive engagement, this research contributes to developing scalable solutions that promote long-term improvements in sleep quality and overall well-being
The SLIIT Research Document Archive (RDA) is the institutional repository of SLIIT, managed by the SLIIT Library. The primary purpose of SLIIT RDA is to manage, store, and disseminate SLIIT research output with its community and beyond, reaching the wider public. This plays a pivotal role in preserving the academic legacy of the institute.
The collection comprises the research output of SLIIT staff and postgraduate research students, including research publications, conference and symposium papers, books, book chapters, theses, and other scholarly materials. Access to full texts may be restricted depending on the access and licensing terms.