Browsing by Author "Rasnayaka, S"

Now showing 1 - 3 of 3

item: Conference-Full-text
Computer vision library for western music sheet notations
(Department of Computer Science and Engineering, University of Moratuwa., 2015-10) Rasnayaka, S; Pemasiri, A; Bandara, M; Meedeniya, D; Perera, I
This paper discusses a computer vision system to detect western music notations from images. The developed library will take in images of western music sheet notation and identify the key features necessary to extract the notes. The images will go through several pre-processing stages and then using straight line detection techniques the staff and notes will be detected. The paper will discuss the algorithms used and developed to achieve this. Finally the paper will present the accuracy measures in the developed system for different types of images.
item: Conference-Abstract
Quantitative and qualitative evaluation of performance and robustness of image stitching algorithms
Dissanayake, V; Herath, S; Rasnayaka, S; Seneviratne, S; Vidanaarachchi, R; Gamage, CD
Many different image stitching algorithms, and mechanisms to assess their quality have been proposed by different research groups in the past decade. However, a comparison across different stitching algorithms and evaluation mechanisms has not been performed before. Our objective is to recognize the best algorithm for panoramic image stitching. We measure the robustness of different algorithms by means of assessing image quality of a set of panoramas. For the evaluation itself, a varied set of assessment criteria are used, and the evaluation is performed over a large range of images captured using differing cameras. In an ideal stitching algorithm, the resulting stitched image should be without visible seams and other noticeable anomalies. An objective evaluation for image quality should give results corresponding to a similar evaluation by the Human Visual System. Our conclusion is that the choice of stitching algorithm is scenario dependent, with run-time and accuracy being the primary considerations.
item: Article-Full-text
Self-supervised vision rransformers for malware setection
(IEEE, 2022) Seneviratne, S; Shariffdeen, R; Rasnayaka, S; Kasthuriarachchi, N
Malware detection plays a crucial role in cyber-security with the increase in malware growth and advancements in cyber-attacks. Previously unseen malware which is not determined by security vendors are often used in these attacks and it is becoming inevitable to find a solution that can self-learn from unlabeled sample data. This paper presents SHERLOCK, a self-supervision based deep learning model to detect malware based on the Vision Transformer (ViT) architecture. SHERLOCK is a novel malware detection method which learns unique features to differentiate malware from benign programs with the use of image-based binary representation. Experimental results using 1.2 million Android applications across a hierarchy of 47 types and 696 families, shows that self-supervised learning can achieve an accuracy of 97% for the binary classification of malware which is higher than existing state-of-the-art techniques. Our proposed model is also able to outperform state-of-the-art techniques for multi-class malware classification of types and family with macro-F1 score of .497 and .491 respectively.