ICITR - 2022

Permanent URI for this collectionhttp://192.248.9.226/handle/123/20685

Browse

Now showing 1 - 20 of 27

item: Conference-Abstract
7th International Conference in Information Technology Research 2022 (Pre Text)
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
item: Conference-Abstract
Polynomial regression real patient state estimate for clinical decision-making
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Hung, CY; Wang, CY; Chen, KW; Yang, CY; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
With the progress of the times, science and technology are changing with each passing day. Clinical decision has become more and more important in medicine nowadays. Clinical decision not only helps clinicians to get immediately crucial decisions; but also provides advices to inexperienced clinicians. In the early days, clinicians could only rely on their own experience and medical reports to make decisions. This process that clinicians analyze patients was very time-consuming. In order to solve these problems, we developed a scoring model. We can analyze patient conditions according to the value of each parameter by using the patient data collected by the hospital. Through computer analysis, evaluations, predictions and optimizations, the suitable model for clinicians and patients can be built. In this paper, we propose a nonlinear polynomial regression approach as a model for predicting patient health scores. The model that predicts patient health score fits multiple researches and clinical examinations through computer simulations. The predicted results are corresponded to the real results when we use the model. With the benefit of the model, it would be easier for clinicians to make clinical decision. In conclusion, our model can not only analyze patient’s conditions, but also predict patient health score via the support of appropriate parameters. This model has the potential to become a valuable tool for clinicians on clinical decision-making in the near future.
item: Conference-Abstract
Analysis and prediction of severity of united states countrywide car accidents based on machine learning techniques
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Boyagoda, LS; Nawarathna, LS; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
The number of vehicles and road transportation increases rapidly daily. Hence the frequency of road accidents and crashes also gradually increase with it. Analyzing traffic accidents is one of the essential concerns in the world. Due to the considerable number of casualties and fatalities caused by those accidents, taking necessary actions to reduce road accidents is a vital public safety concern and challenge worldwide. Various statistical methods and techniques are used to address this issue. Hence, those statistical implementations are used for multiple applications, such as extracting cause and effect to predict real-time accidents. In this study, a United States (US) Countrywide car accidents data set consisting of about 1.5 million accident records with other relevant 45 measurements related to the US Countrywide Traffic Accidents were used. This work aims to develop classification models that predict the likelihood of an accident is severe. In addition, this study also consists of descriptive analysis to recognize the key features affecting the accident severity. Supervised machine learning methods such as Decision tree, K-nearest neighbour, and Random forest were used to create classification models. The predictive model results show that the Random Forest model performs with an accuracy of 83.95% for the train set and 80.69% for the test set, proving that the Random forest model performs better in accurately detecting the most relevant factors describing a road accident severity.
item: Conference-Abstract
Walkable space estimation for visually impaired persons in 3d simulated environment
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Silva, CS; Wimalaratne, P; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Environmental perception is the process by which visually impaired people tend to sense, understand, and build awareness of the walking environment and the surrounding objects. Walkable space includes any space that the visually impaired person is physically capable of walking. The task of estimating the walkable space is performed by estimating pixels belonging to the ground plane in the scene. Plane parameters are estimated based on the input of 3D point coordinates of pixels belonging to the sidewalk in the outdoor environment. Moreover, the curb line is estimated using the output of the walkable space calculation to determine whether the visually impaired navigator is walking along the sidewalks or on the road. Proof of concept and evaluation experiments are conducted on a 3D simulated environment with the use of RGB, semantic and depth images taken from the wearable RGB camera, segmentation camera and depth camera. The results are benchmarked with the existing image processing-based methods and showed that the proposed method can be successfully implemented for the safe navigation of visually impaired users with minimum computational complexity in image processing.
item: Conference-Abstract
Automating the initial configuration of sdn switches in a hybrid-sdn environment
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Bolonghe, WKN; Rupasinghe, PM; Umayanga, KSB; Weerasiri, KLHI; De Silva, D; Wijesiri, P; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
At present for SDN environments, there are many proposed mechanisms to improve its reliability, and performance. The challenges faced by SDN networks are mainly related to Scalability, Reliability and Performance. As an example, for scalability related challenges, many SDN networks have problems when replacing or installing new SDN switches to the network. The problems are mainly the cost and errors while the installation process. To overcome this issue, a mechanism is proposed to automate the initial configuration of the newly added SDN switches. When it comes to the performance, there are problems such as load balancing, looping and traffic related issues due to broadcast messages. To minimize these challenges the project also brings solutions by implementing DHCP relays and STP inside the network. Also, with the multi-controller architecture proposed it increase the efficiency and Performance. To automate the initial configuration of the switches, the newly added switch is detected at first and then a relationship is established between the newly added switch and an existing SDN switch. Then the newly added switch establishes a connection with the automation server, which then results the automation process to start. The proposed mechanism is implemented in a testing environment using Mininet which creates a virtual environment, a RYU controller as the SDN controller and the OpenFlow is used as the protocol for communication between interfaces. The proposed mechanism will bring many benefits like minimize the errors and save time due to the automated initial configuration in a hybrid-SDN environment, increase efficiency and Performance comparing to the present hybrid-SDN environments.
item: Conference-Abstract
Gender and age estimation from facial images using deep learning
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Thaneeshan, R; Thanikasalam, K; Pinidiyaarachchi, A; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Automated gender and age estimation from facial images are important for many realworld applications. Although, several studies have been proposed in the past, most of them are proposed as individual models and a considerable performance gap is noticed. Moreover, deep learning based approaches treated their model as a black box classifier and hence their model’s knowledge representation is not understandable and difficult to further improve. In this manuscript, we have proposed a simple and efficient CNN model architecture by considering gender and age estimation as a multi-label classification problem. The proposed model is trained and then evaluated on the publicly available Adience benchmark dataset. Experimental results demonstrated that the proposed model showed better performance than the similar approaches with an accuracy of 84.20 % on gender estimation and an accuracy of 57.60 % on age estimation. In addition, we have proposed a visualization technique to explain the classification results and then the gender-specific and age group-specific landmark facial regions are identified.
item: Conference-Abstract
Automated commentary generation based on fps gameplay analysis
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Mamoru, DLS; Panditha, AD; Perera, ASSJ; Ganegoda, GU; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Video games evolved into competitive sports within a short span of time. Commentary plays a key role in any sport. Commentary is useful to understand the game as well as to capture key moments of the game. The balance between play-to-play commentary and color commentary creates the whole value of commentary. Additionally, it has been proven that contribution factors play a major role in the meaning of an automated commentary as well. The proposed commentary generator takes all three factors into account to generate a commentary track for the first-person shooter game Valorant. Human evaluation along with numerical figures are taken to evaluate the quality of the generated commentary. The evaluation results suggest that the proposed model performs on par with the human commentary.
item: Conference-Abstract
Using CNN to identify the condition of edible fish
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Mahendran, R; Seneviratne, GP; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Evaluating the edibility of fish by it’s freshness is an essential process for the fisheries industry as it contributes to customers’ health and the taste of food. In general, identifying freshness of fish is a difficult task for the customers due to lack of experience or knowledge. Using a real-time application which employs real-time images of fish is the best solution to identify their freshness. In this study, a model was developed using VGG16 architecture in a deep convolutional neural network (CNN) to extract the features of the fish and to classify them based on their freshness. Here, Bluefin Trevally fish was selected as a sample and its freshness was detected using real-time images. Those images were collected in various backgrounds with different lightning by different devices. In real-time images, features of fish such as the colour of the eye and frozen blood colour of the operculum were used to identify the freshness of fish. An accuracy of 99% on identification of freshness was achieved by this model.
item: Conference-Abstract
A dual cnn architecture for single image raindrop and rain streak removal
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Sivaanpu, A; Thanikasalam, K; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Visual quality of rainy images are considerably poor due to the raindrops in camera lens and the rain streaks in the background scenes. Although the raindrops and rain streaks are appeared together in real-world rainy images, most of the previous approaches are proposed to remove either of them. In this paper, we have proposed a novel CNN model architecture to remove raindrops and rain streaks together. The proposed CNN model architecture has two branches and it consumes two formats of a rainy image via an encoder-decoder network and a dense CNN network. At the end of the architecture, outputs of both branches are combined to produce a highvisibility rain free image with natural colors. In addition, internal and external skip connections are introduced in the blocks of these branches to improve the performance further. The proposed model is trained and then tested on Raindrop, Rain100H, Rain100L, and Rain12 benchmarks and showed excellent performance than the state-of-the-art approaches.
item: Conference-Abstract
Eduzone – a educational video summarizer and digital human assistant for effective learning
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Wangchen, T; Tharindi, PN; Chaveena De Silva, KCC; Sandeepa, WD; Kodagoda, N; Suriyawansa, K; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
The availability of technology and the expansive nature of the internet have created a surge in the demand for online learning. Despite so many advantages, there are some existing drawbacks related to online learning. The lengthy recorded video lectures of different subjects and modules in a static manner, are extremely tedious for the learner to understand the contents available. And lack of assistance for academic-related problems of students is also stated as a major issue that comes with online education. EDUZONE provides a reliable solution to mitigate and overcome these challenges. This tool is educational assistance that generates a summarized version of the video lectures which depicts the overall idea of the whole video with the capability of a lecture notes generator along with a digital human which helps to clarify students’ problems and build an efficient conversational flow. The summarized video content can be used by the learners for revisions and as a quick reference before any examinations. In addition to generating short and precise content, EDUZONE also indexes any specific topics to make it easier to find content and generate class notes, highlighting all the important content. Overall EDUZONE can be considered a time-efficient educational assistant which helps students with their studies.
item: Conference-Abstract
Multi-ai based wireless sensor node for forest fire detection
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Pieris, TPD; Kulasekera, AL; Chathuranga, KVDS; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Forest fires are one of the major environmental issues in many countries around the world. There are many ways to detect these forest fires, such as optical camera-based systems, satellite-based systems, fire prediction systems, and wireless sensor nodes (WSN). They have major issues such as line-of-sight limitations in optical camera-based systems, low detection delay of satellite systems, and need more time to model the environments for fire prediction systems. The most suitable way of fire detection is the usage of wireless sensor network based systems because they have low detection delay, more fire behavioral information, good stability, etc. In WSNs, there are many methods of fire detection such as thresholding data, fuzzy logic, neural networks, multi-AI, etc. Here, the proposed WSN based system consists of multi-AI. The system was evaluated with test fires. The proposed system could successfully identify the created fires, and this system is simpler than existing systems.
item: Conference-Abstract
Closed-loop dc motor embedded control platform based on arm® for distance learning experiments
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Ganganath, R; Ranganath, C; Jayasekara, DC; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
DC motor controlling is known to be one of the most crucial application in the Engineering world, where precise motion control is commonly found in industrial and commercial applications. Hence, learning the fundamentals of DC closed-loop motor control is beneficial for undergraduate students studying Engineering courses. However Due to the ongoing unprecedented economic crisis in Sri Lanka, the lack of access to electronic components, has become very challenging and in turn, affects implementing and testing of the theories on real prototypes. Hence In this paper, a low-cost ARM® based DC motor embedded control platform is presented so the students who are following control courses can implement their own prototype for an affordable cost using only a few common components. The closed-loop control algorithm is designed by using the PID controller, tuned the system using the classical: trial & error method and tested the system responses in real-time, according to different load conditions. Moreover, ARM® based STM®32F407G micro-controller has been used to program the controller and keil® uVision® software is used as the programming IDE. Using MATLAB® and Simulink® platforms, a comparison of the system responses with no-load, lite load and full-load conditions has been presented.
item: Conference-Abstract
Intelligent wheelchair with emotion analysis and voice recognition
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Perera, S; Gamage, S; Weerasinghe, C; Jayawardena, C; Pathinayake, K; Rajapaksha, S; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Intelligent wheelchairs are becoming more and more prevalent in contemporary life, and the peaceful interaction of humans with wheelchairs is one of the most popular research topics. The development of a voice recognition and emotion recognition based intelligent wheelchair framework is being addressed here for truly impaired/disabled people who are unable to operate the wheelchair by hand. The patient can operate the wheelchair using voice commands, and the wheelchair’s Emotion Analysis module recognizes the patient’s face and records the patient’s emotions before sending the information to a cell phone application. A portion of the intelligent wheelchair is made to gather crucial information given by other units and send out emergency calls or notifications to the caregivers. Face recognition technology uses image processing to identify facial expressions by detecting the patient’s face and facial expressions. This helps the other components collect and send data via Internet of Things technologies. Speech – to –Text and Text – to- Speech Methodology is used in the voice recognition module and it captures the voice command data set and extracts the features of the commands. The model is already built and trained to recognize the commands and to send action request to the relevant unit. The Responsive AI auto starts the timer when the patient moves away from the wheelchair, recognizes time and responses back. This unit auto also sends the alert and calls to the guardian when the user has no response.
item: Conference-Abstract
Design and testing of an arduino-based network jammer device
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Najath, MNM; Herath, HMDS; Rajapakse, A; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
A computer network is described as two or more devices connecting and communicating, each known as a computer network. These types of networks work on wired and wireless mediums. A sort of transmission medium is wired communications, also referred to as directed media. This form of communication is more reliable than wireless since it is the most stable. In wired communication, twisted pair cable, coaxial cable, and fiber optic cable are used to send data from the source to the destination. Wireless communication is also called "unguided" or "unrestricted" transmission media. In this mode, no physical medium is required to transmit electromagnetic signals. On the other hand, in wireless communication, a message can transfer through the air, water, or vacuum, i.e., infrared, radio, and microwave waves. Therefore, this research mainly focuses on wireless communication using radio frequency technology. This technology is also used on TVs, radios, laptops, modems, mobile phones, and pagers. People constantly interact with this technology in the modern world, and all these people have secure data and confidential information. However, in this case, someone uses this technology and tries to get the data illegally. It is a big problem for all civilians and the government. In this situation, a network jammer device can be used to protect those data from snooping attempts, location tracking, limiting cell phone activity, and cheating. This paper describes a low-cost network jammer that uses an Arduino and an RF 433 MHz module to create a frequency between 315 MHz and 433 MHz. It uses the denial-of-service technique to prevent illegal activities within a 100-meter range.
item: Conference-Abstract
Development of nonlinear dynamics simulator for 2-dof ball balancing platform to assist distance learning of control systems
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Ranganath, C; Annasiwaththa, B; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
In a wide range of control applications, balance control systems are known to be one of the most crucial and challenging applications. 2-DOF ball balancing platform is an experimental platform which can perform various kinds of balance control experiments. Hence, this laboratory experiment is widely used in many control engineering courses. But with COVID- 19, travel restrictions and social distancing, performing laboratory experiments have become much more challenging. Hence, this research explores the feasibility of performing control engineering experiments on Ball on plate system in a simulation environment for distance learning purposes. This paper represents the control design theory, system modelling & simulations of ball balancing platform and proposes a method of performing lab experiments for remote-learning. All the system modelling and simulations of this research were done using MATLAB Simulink based on the Sims cape model developed using SolidWorks.
item: Conference-Abstract
Computer vision based navigation robot
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Haputhanthri, M; Himasha, C; Balasooriya, H; Herath, M; Rajapaksha, S; Harshanath, SMB; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
The majority of industrial environments and homewares need help when exploring unknown locations owing to a lack of understanding about the building structure and the various impediments that may be faced while transporting products from one spot to another. This is because there is a lack of knowledge about the building structure and the potential obstacles that may be encountered. This paper provides “Computer Vision-Based Navigation Robot” as a strategy for indoor navigation with optimal accessibility, usability, and security, decreasing issues that the user may encounter when traveling through indoor and outdoor areas with real-time monitoring of the most up to date IoT technology. The article is titled “Indoor Navigation with Optimal Accessibility, Usability, and Security.” This article proposes “Computer Vision-Based Navigation Robot” as a solution for interior navigation that provides optimum accessibility, usability, and security. This is done in order to tackle the issue that was presented before. Since the readers of this post include people who work in industry as well as physically challenged people who live alone, CVBN Robot takes object-based inputs from its surroundings. This is because the audience for this essay includes both groups of people. This study also covers a variety of methods for localization, sensors for the detection of obstacles, and a protocol for an Internet of Things connection between the server and the robot. This connection enables real-time position and status updates for the robot as it navigates a known but unknown interior environment. In addition, this study covers a variety of methods for localization, sensors for the detection of obstacles, and a protocol for an Internet of Things connection between the server and the robot.
item: Conference-Abstract
Elderly care home robot using emotion recognition, voice recognition and medicine scheduling
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Kularatne, BMUS; Basnayake, BMJN; Sathmini, PDLAM; Sewwandi, GVU; Rajapaksha, S; De Silva, D; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
The robotic concept is used for several tasks to easier human day-to-day tasks. There are various recreational studies have been done on the elderly people’s care system. In this study, the system can identify the elderly people’s emotional status using thermal image processing that eliminates the halo effect issue in thermal images using a single discriminator Cycle-GAN model, serving medicine to elderly people by moving towards the elderly person while avoiding obstacles using point to point algorithm and obstacle avoidance and identify the semantic analysis by using web ontology based language. The integrated system is evaluated using the Gazebo simulator because the cost is lower than implementing the features in a real robot.
item: Conference-Abstract
Elderly care home robot using emotion recognition, voice recognition and medicine scheduling
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Kularatne, BMUS; Basnayake, BMJN; Sathmini, PDLAM; Sewwandi, GVU; Rajapaksha, S; De Silva, D; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
The robotic concept is used for several tasks to easier human day-to-day tasks. There are various recreational studies have been done on the elderly people’s care system. In this study, the system can identify the elderly people’s emotional status using thermal image processing that eliminates the halo effect issue in thermal images using a single discriminator Cycle-GAN model, serving medicine to elderly people by moving towards the elderly person while avoiding obstacles using point to point algorithm and obstacle avoidance and identify the semantic analysis by using web ontology based language. The integrated system is evaluated using the Gazebo simulator because the cost is lower than implementing the features in a real robot.
item: Conference-Abstract
Peer learning – an interactive and collaborative elearning application for college student
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Sirithunga, HAPM; Deshan, BGS; Sigera, PHD; Udagedara, PY; Samarakoon, U; Kumari, S; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
Since the start of the COVID-19 epidemic in 2020, the entire educational system has been challenging and Sri Lanka economic crisis, but this is especially effect for students who are now enrolled. This developmental milestone is reached when adolescents begin to assume responsibilities and acquire leadership skills through participation in a range of team activities. It is easiest to gain experience working in a group setting while still in school. Nevertheless, given the current stage of the Sri Lanka economic crisis, students will face a range of challenges. They are incapable of participating in group activities that are relevant to the subjects they teach, and, as previously indicated, enhancing their leadership skills, which is particularly problematic when working with students. The "Peer Learning" solution is a web-based application that supports students in enhancing their collaborative learning skills. Through the system, students have the opportunity to study a variety of collaborative tasks, which improves their educational and interpersonal abilities. In addition, professors can share their knowledge with students by personalizing questions, posting films, and demonstrating figures. Students can easily comprehend the system's operation due to its user- friendly design, which enables advanced technological methods for monitoring and guiding students' activities simultaneously.
item: Conference-Abstract
A visually interpretable forensic deepfake detection tool using anchors
(Information Technology Research Unit, Faculty of Information Technology, University of Moratuwa., 2022-12) Jayakumar, K; Skandhakumar, N; Sumathipala, KASN; Ganegoda, GU; Piyathilake, ITS; Manawadu, IN
“Deepfakes” have seen a dramatic rise in recent times and are becoming quite realistic and indistinguishable with the advancement of deepfake generation techniques. Promising strides have been made in the deepfake detection area even though it is a relatively new research domain. Majority of current deepfake detection solutions only classify a video as a deepfake without providing any explanations behind the prediction. However, these works fail in situations where transparency behind a tool’s decision is crucial, especially in a court of law, where digital forensic investigators maybe called to testify if a video is a deepfake with evidence; or where justifications behind tool decisions plays a key role in the jury’s verdict. Explainable AI (XAI) has the power to make deepfake detection more meaningful, as it can effectively help explain why the detection tool classified the video as a deepfake by highlighting forged super-pixels of the video frames. This paper proposes the use of “Anchors” XAI method, a model-agnostic high precision explainer to build the prediction explainer model, that can visually explain the predictions of a deepfake detector model built on top of the EfficientNet architecture. Evaluation results show that Anchors fair better than LIME in terms of producing visually explainable and easily interpretable explanations and produces an anchor affinity score of 70.23%. The deepfake detector model yields an accuracy of 91.92%.

Browse

Recent Submissions