Browsing by Author "Sotheeswaran, S."

Now showing 1 - 3 of 3

Automatic road traffic signs detection and recognition using ‘You Only Look Once’ version 4 (YOLOv4)
(Department of Industrial Management, Faculty of Science, University of Kelaniya Sri Lanka, 2021) Fernando, W. H. D.; Sotheeswaran, S.
Using scenario transformation methodology, we identified four scenarios that indicated a lack of trusted parties to sell harvest has forced smallholder farmers to sell the harvest to brokers who often collect the harvest at the farm gate at the lowest possible prices and sell in the market for large profits. As blockchain smart contracts provide a mechanism to reduce risk and establish trust between unknown trading partners, we transformed these into a scenario that establishes trust between farmer and unknown broker using smart contracts, generating a trust-enabled market. This scenario enables farmers to search for the optimum farm-gate price without relying on known brokers. The scenario is further enhanced to enable a Many-one-Many market linkage, facilitating automatic aggregated marketing. The paper presents the functional prototype of the scenario, explaining the functionality of the transformed system.
Novel computational approaches for border irregularity prediction to detect melanoma in skin lesions
(Department of Industrial Management, Faculty of Science, University of Kelaniya, Sri Lanka, 2020) Abeysinghe, D.V.D.S.; Sotheeswaran, S.
Medical image detection has been a rapidly growing field of study during the last few years. There are different challenges associated with it. Many works have been done in order to provide solutions for key challenges. This study of work is focused on melanoma detection by using Asymmetry, Border irregularity, Colour textures, and Diameter (ABCD) feature along with proposing two new approaches for border irregularity detection. The proposed two new approaches are distance difference method and gradient method, which follows the main concept as traversing along the continuous borderline of the lesion. Further, this study varies from the existing studies, since it has been taken counts of distances from the centroid to the borderline without considering the distance from the image border to the borderline of the lesion. It was able to achieve a classification rate of 79% and 78.5% using distance difference method and gradient method, respectively whereas the classification without the border irregularity feature achieved 78% of accuracy performing on PH2 dataset. Further, this study can be stated as most appropriate to classify non-melanoma rather than melanoma. It is contributed by generating simple computer science-based approaches rather than complex mathematical methods to detect border irregularity and makes the medical image detection easy.
A tree structure-based classification of diabetic retinopathy stages using convolutional neural network
(Department of Industrial Management, Faculty of Science, University of Kelaniya Sri Lanka, 2021) Peiris, M. S. H.; Sotheeswaran, S.
Detection, and classification of medical images have become a trending field of study during the last few decades. There is a considerable amount of vital challenges to be overcome. Ample work has been carried out to provide proper solutions for those key challenges. This study was carried out to extend one such medical image classification process to classify the stages of Diabetic Retinopathy (DR) images from colour fundus images. The study proposes a novel Convolutional Neural Network (CNN) architecture which is considered to be one of the most trending and efficient forms of classification of DR stages. Initially, the pre-processing techniques were employed to the DR fundus images with Green channel extraction and Contrast Limited Adaptive Histogram Equalization (CLAHE). The data augmentation strategy was utilised to increase training images from the DR images. Finally, Feature extraction and classification were carried out by using the proposed CNN architecture. It consists of a 14 layered CNN model, which continues three main classifications. In this proposed classification, the images were classified into a tree structure based binary classification as No_DR and DR at the beginning, and then the DR images were again classified into two classes, namely Pre_Intermediate and Post_Intermediate. Moreover, those two classes were again separately classified into Mild, Moderate, and Proliferate_DR, Severe, respectively. The Kaggle is one of the benchmark dataset repositories which was used in this study. The proposed model was able to achieve accuracies of 81%, 96%, 84%, and 97% for the above-mentioned classifications, respectively.