Research Papers - Dept of Software Engineering

Permanent URI for this collectionhttps://rda.sliit.lk/handle/123456789/1022

Browse

Filters

2011 - 2019352020 - 202537

Advanced Search

Filter by

Settings

Author: search.filters.author.Kasthurirathna, DHas files: Yes

Search Results

Now showing 1 - 10 of 72
  • Thumbnail Image
    PublicationEmbargo
    Spatio-temporal graph neural network based child action recognition using data-efficient methods: A systematic analysis
    (Elsevier Inc, 2025-06-03) Mohottala, S; Gawesha, A; Kasthurirathna, D; Samarasinghe, P; Abhayaratne, C
    This paper presents implementations on child activity recognition (CAR) using spatial–temporal graph neural network (ST-GNN)-based deep learning models with the skeleton modality. Prior implementations in this domain have predominantly utilized CNN, LSTM, and other methods, despite the superior performance potential of graph neural networks. To the best of our knowledge, this study is the first to use an ST-GNN model for child activity recognition employing both in-the-lab, in-the-wild, and in-the-deployment skeleton data. To overcome the challenges posed by small publicly available child action datasets, transfer learning methods such as feature extraction and fine-tuning were applied to enhance model performance. As a principal contribution, we developed an ST-GNN-based skeleton modality model that, despite using a relatively small child action dataset, achieved superior performance (94.81%) compared to implementations trained on a significantly larger (x10) adult action dataset (90.6%) for a similar subset of actions. With ST-GCN-based feature extraction and fine-tuning methods, accuracy improved by 10%–40% compared to vanilla implementations, achieving a maximum accuracy of 94.81%. Additionally, implementations with other ST-GNN models demonstrated further accuracy improvements of 15%–45% over the ST-GCN baseline. The results on activity datasets empirically demonstrate that class diversity, dataset size, and careful selection of pre-training datasets significantly enhance accuracy. In-the-wild and in-the-deployment implementations confirm the real-world applicability of above approaches, with the ST-GNN model achieving 11 FPS on streaming data. Finally, preliminary evidence on the impact of graph expressivity and graph rewiring on accuracy of small dataset-based models is provided, outlining potential directions for future research. The codes are available at https://github.com/sankamohotttala/ST_GNN_HAR_DEML.
  • Thumbnail Image
    PublicationEmbargo
    DS-HPE: Deep Set for Head Pose Estimation
    (IEEE, 2023-04-18) Menan, V; Gawesha, A; Samarasinghe, p; Kasthurirathna, D
    Head pose estimation is a critical task that is fundamental to a variety of real-world applications, such as virtual and augmented reality, as well as human behavior analysis. In the past, facial landmark-based methods were the dominant approach to head pose estimation. However, recent research has demonstrated the effectiveness of landmark-free methods, which have achieved state-of-the-art (SOTA) results. In this study, we utilize the Deep Set architecture for the first time in the domain of head pose estimation. Deep Set is a specialized architecture that works on a “set” of data as a result of the “permutation-invariance” operator being utilized in the model. As a result, the model is a simple yet powerful and edge-computation-friendly method for estimating head pose. We evaluate our proposed method on two benchmark data sets, and we compare our method against SOTA methods on a challenging video-based data set. Our results indicate that our proposed method not only achieves comparable accuracy to these SOTA methods but also requires less computational time. Furthermore, the simplicity of our proposed method allows for its deployment in resource-constrained environments without the need for expensive hardware such as Graphics Processing Units (GPUs). This work underscores the importance of accurate and resource-efficient head pose estimation in the fields of computer vision and human-computer interaction, and the Deep Set architecture presents a promising approach to achieving this goal.
  • Thumbnail Image
    PublicationEmbargo
    Expert System for Kubernetes Cluster Autoscaling and Resource Management
    (IEEE, 2022-12-09) Hettiarachchi, L.S; Jayadeva, S.V; Bandara, R. A. V; Palliyaguruge, D; Samaratunge Arachchillage, U. S. S; Kasthurirathna, D
    The importance of orchestration tools such as Kubernetes has become paramount with the popularity of software architectural styles such as microservices. Furthermore, advancements in containerization technologies such as Docker has also played a vital role when it comes to advancements in the field of DevOps, enabling developers and system engineers to deploy are manage applications much more effectively. However, infrastructure configuration and management of resources are still challenging due to the disjointed nature of the infrastructure and resource management tools’ failure to comprehend the deployed applications and create a holistic view of the services. This is partly due to the extensive knowledge required to operate these tools or due to the inability to perform specific tasks. As a result, multiple tools and platforms need to conFigure together to automate the deployment, monitoring and management processes to provide the optimal deployment strategy for the applications. In response to this issue, this research proposes an expert system that creates a centralized approach to cluster autoscaling and resource management, which also provides an automated low-latency container management system and resiliency evaluation for dynamic systems. Furthermore, the time series load prediction is done using a BiLSTM and periodically creates an optimized autoscaling policy for cluster performance, thus creating a seamless pipeline from deployment, monitoring scaling, and troubleshooting of distributed applications based on Kubernetes.
  • Thumbnail Image
    PublicationEmbargo
    SPAVIS: Mobile Application for Visually Impaired Based on Assistive Software and Volunteerism
    (IEEE, 2022-12-09) Mahir, M.A.M.; Hussain, M.N.M.; Perera, R.D.D; Upendra, Y.A.M; Wickramarathne, C. J.; Kasthurirathna, D
    Sri Lankan population accounts up to almost one million visually impaired individuals out of which are mostly students and young individuals. As the educational structure for the visually impaired improves with funds, blind schools, and free education, assistance with minute needs for most visually impaired individuals comes at a cost. There are many assistive technologies, such as audio books, screen magnifiers, braille books, and screen readers, prevalent around the island. However, there are several limitations to these technologies, mainly their availability and affordability. In Sri Lanka, many individuals, societies, clubs, and many more are willing to volunteer to help those in need, even those that require physical attention. As much as it is anticipated to aid those in need, there is very little attention to the ways it can be done. Hence, this research provides a way to develop a user-friendly mobile application with assistive software and volunteerism to aid visually impaired students with their daily needs.
  • Thumbnail Image
    PublicationEmbargo
    'xīnl' The Social Media App to Replenish Mental Health with the Aid of an Egocentric Network
    (Institute of Electrical and Electronics Engineers, 2022-11-03) Kasthurirathna, D; Kalansooriya, S; Kaluarachchi, A; Weerawickrama, C; Nanayakkara, D; Adeepa, D
    The impact of social groups on one's emotional health is a crucial issue that must be addressed correctly. Emotions and social groups play significant roles in human mental and physical activities. It is difficult to detect and maintain track of changing emotional states. The main goal of this study is to build a social media app called Xinli, that proposes an aggregated method to predict emotions using a multimodal approach and to predict personalized activities based on the user's mental state, and to further track the improvement of emotional state with the impact of recommended activities and social support groups. The results suggest that the aggregated modalities method is more accurate in recognizing emotions, and activity prediction using reinforcement learning is a clean way to personalize activities based on the emotional state from user to user, which is the novelty of the proposed study.
  • Thumbnail Image
    PublicationEmbargo
    Better you: Automated tool that evaluates mental health and provides guidance for university students
    (Institute of Electrical and Electronics Engineers Inc., 2022-11-04) Eeswar, S. S; Samaratunga, J. S; Nivethika, G; Anjana, W.W.M.; Jayasingha, T.B.; Pandithakoralage, S; Kasthurirathna, D
    This research paper proposes a system that evaluates mental health through text-based, voice-based and facial emotion recognition. After predicting the user's overall emotional state, activity suggestions and close contact interactions will be suggested to improve their mental health.
  • Thumbnail Image
    PublicationOpen Access
    Disassortative Mixing and Systemic Rational Behaviour: How System Rationality Is Influenced by Topology and Placement in Networked Systems
    (MDPI, 2022-09-12) Kasthurirathna, D; Ratnayake, P; Piraveenan, M
    Interdependent decisionmaking of individuals in social systems can be modelled by games played on complex networks. Players in such systems have bounded rationality, which influences the computation of equilibrium solutions. It has been shown that the ‘system rationality’, which indicates the overall rationality of a network of players, may play a key role in the emergence of scale-free or core-periphery topologies in real-world networks. In this work, we identify optimal topologies and mixing patterns of players which can maximise system rationality. Based on simulation results, we show that irrespective of the placement of nodes with higher rationality, it is the disassortative mixing of node rationality that helps to maximize system rationality in a population. In other words, the findings of this work indicate that the overall rationality of a population may improve when more players with non-similar individual rationality levels interact with each other. We identify particular topologies such as the core-periphery topology, which facilitates the optimisation of system rationality. The findings presented in this work may have useful interpretations and applications in socio-economic systems for maximizing the utility of interactions in a population of strategic players.
  • Thumbnail Image
    PublicationEmbargo
    SMART Garbage Bin Kit Expandable and Intelligent Waste Management System using Deep Learning and IoT for Modern Organizations
    (IEEE, 2021-12-02) Hewagamage, P.; Perera, D; Thilakarathna, T; Kasthurirathna, D; Fernando, R; Mihiranga, A
    According to published statistics, Sri Lanka produces garbage around 7000MT per day, and every organization directly contributes this national amount depending on the waste management practices. 'Waste contamination' is a critical issue that affects waste management, and it should be addressed during the garbage collection process. This has led to environmental hazards resulting in health and other social issues. Hence, it is a responsibility of an organization to separate the garbage during the collection process using a suitable technique. In this paper, we are proposing a smart garbage bin kit that automates the separation of garbage collection, which minimizes human error using AI-based technologies. IoT-based devices connected to a smart garbage bin kit guide the user to the correct bin. At the same time, our proposed system can be easily expanded for new special waste categories as well. The other important issue of the current garbage management is improper time management of the garbage removal process in organizations. This happens due to the lack of real-time data on waste bins, and collection is based on the fixed time interval irrespective of the status and location of garbage bins. In the proposed system of SMART Garbage Bin Kit, the group of all interconnected garbage bins is monitored in real-time to identify the optimum collection path considering the location and the status of garbage bins using an optimized algorithm. Hence, the study presented in this paper integrates several intelligent approaches together with IoT based network to build a cutting-edge device, declared as SMART Garbage Bin kit. The prototype system has been built as a part of the research study to demonstrate its feasibility and sustainability.
  • Thumbnail Image
    PublicationOpen Access
    A Singlish Supported Post Recommendation Approach for Social Media
    (SCITEPRESS – Science and Technology Publications, 2022-01) Sandamini, U; Rathnakumara, K; Pramuditha, p; Dissanayake, M; Sriyaratna, D; De Silva, H; Kasthurirathna, D
    Social media is an attractive means of communication which people used to exchange information. Post recommendation eliminates the overflooding of information in social media to the users’ news feed by suggesting the best matching information based on users’ preference that in return increase the usability. Social media users use different languages and their variations where most of the Sri Lankan users are accustomed to use Sinhala and Romanized Sinhala. However, post recommendation approaches used in current social media applications do not cater to code-mixed text. Therefore, this paper proposes a novel post recommendation approach that supports Singlish. The study is separated into two major components as language identification and transliteration, and post recommendation. In this study, script identification was performed using regular expressions while a Naïve Bayes classification model that accomplished 97% of accuracy was employed for language identification of Romanized text. Transliteration of Singlish to Sinhala was conducted using a character level seq2seq BLSTM model with a BLEU score of 0.94. Furthermore, Google translation API and YAKE were used for Sinhala-English translation and keyword extraction respectively. Post recommendation model utilized a combination of rule-based and CF techniques that accomplished the RMSE of 0.2971 and MAE of 0.2304.
  • Thumbnail Image
    PublicationOpen Access
    Disassortative mixing of boundedly-rational players in socio-ecological systems
    (researchgate.net, 2022-03-25) Ratnayake, P; Kasthurirathna, D; Piraveenan, M
    Bounded rationality refers to the non-optimal rationality of players in non-cooperative games. In a networked game, the bounded rationality of players may be heterogeneous and spatially distributed. It has been shown that the ‘system rationality’, which indicates the overall rationality of a network of players, may play a key role in the emergence of scale-free or core-periphery topologies in real-world networks. On the other hand, scalar-assortativity is a metric used to quantify the assortative mixing of nodes with respect to a given scalar attribute. In this work, we observe the effect of node rationality-based scalar-assortativity, on the system rationality of a network. Based on simulation results, we show that irrespective of the placement of nodes with higher rationality, it is the disassortative mixing of node rationality that helps to maximize system rationality in a population. The findings may have useful interpretations and applications in socio-economic systems in maximizing the utility of interactions in a population of strategic players