Projects

I'm interested in building resource efficient intelligent systems, mainly performing intelligent tasks on image data such as medical imaging or video recordings. Here are some of the projects we've been working on.

Medical Image Analysis and Diagnostic Asistance

We work on image data such as X-Ray, CT, Electron Microscopy as well as Dermoscopic images. Our efforts are to create diagnostic assistive tools using computer visions methods that can analyze above types of images so that an informative judgement can be made by medical practitioners using the output of such methods. Following are some of our publications in this area

  • Dasanayaka S, Shantha V, Silva S, Meedeniya D, Ambegoda T. Interpretable machine learning for brain tumour analysis using MRI and whole slide images. Software Impacts. 2022 Aug 1;13:100340. [Article]
  • Hussaindeen A, Iqbal S, Ambegoda TD. Multi-label prototype based interpretable machine learning for melanoma detection. International Journal Of Advances In Signal And Image Sciences. 2022 Jan 1;8(1):40-53. [Article]

Satellite Image Analysis / Remote Sensing

In our research on satellite image analysis using computer vision, we explore Mars's surface to better understand its geology and potential for past life. One project involves segmenting serpentine zones on Mars, which helps in studying the planet's geological history. Another focuses on automatically detecting and segmenting non-inverted channels from satellite data, shedding light on Mars’s ancient water flow. These efforts aim to provide insights into Mars's environment and support future exploration

  • Kavinda KP, Rathnayaka CB, Silva WG, Ambegoda TD, Manogaran R, Karunatillake S. RESIST: Tool to Automatically Segment Martian Inverted Channels in HiRISE Images. LPI Contributions. 2023 Mar;2806:1821. [Article]
  • Jayakody DR, Ambegoda TD, Karunatillake S, Hughes EB. Optimized Field Sampling of Mars-Analog Serpentine Zones via Machine Learning. LPI Contributions. 2023 Mar;2806:2242. [Article]

Sensors and Data Science

There are a variety of sensors to capture events in the environment including machines and humans. These could be IMU sensors to detect acceleration, lidar sensors to sense the 3D environment and its changes, wifi transceivers that can capture certain changes in the environment, and vibration sensors that can capture subtle changes in machine vibrations when they are in operation . Our efforts are to develop and apply data science techniques to extract rich features and use them to build tools to support decision making for humans as well as autonomous agents such as robots
  • Schmidt T, Ambegoda T, Gunasekera K. Vehicle Classification Using Raspberry Pi: A Guide to Capturing WiFi CSI Data. In2023 Moratuwa Engineering Research Conference (MERCon) 2023 Nov 9 (pp. 702-707). IEEE. [Article]
  • Wijesinghe A, Ambegoda TD, Perera AS, Sivakumar T. Siamese networks for RF-based vehicle trajectory prediction. In2023 IEEE 17th International Conference on Industrial and Information Systems (ICIIS) 2023 Aug 25 (pp. 495-500). IEEE. [Article]

Sign Language Recognition

Our aim is to build an automatic realtime sign language recognition system for the Sinhala Sign Language that's used in Sri Lanka. We have developed tools to recognize the Sinhala Sign Alphabet as well as a number of words. Currently we are trying to expand the vocabulary as well as the capability to translate sentences in Sinhala Sign Language in realtime. We have already contributed the largest dataset for this purpose and currently working on expanding it further to support similar research and development efforts.
  • Charuka K, Wickramanayake S, Ambegoda TD, Madhushan P, Wijesooriya D. Sign Language Recognition for Low Resource Languages Using Few Shot Learning. InInternational Conference on Neural Information Processing 2023 Nov 20 (pp. 203-214). Singapore: Springer Nature Singapore. [Article]
  • Weerasooriya AA, Ambegoda TD. Sinhala Fingerspelling Sign Language Recognition with Computer Vision. In2022 Moratuwa Engineering Research Conference (MERCon) 2022 Jul 27 (pp. 1-6). IEEE. [Article]

Smart Agriculture

IoT and ML methods have a significant scope to make crop monitoring efficient and effective. We have been working on developing cost effective tools and frameworks to monitor the environment of crops and their growth rate with sensor networks and computer vision methods.
  • Senevirathne I, Ambegoda T, Wijesena R, Perera I. IoT-based Soil Nutrient Analyser using Gaussian Process Regression. In2022 2nd International Conference on Advanced Research in Computing (ICARC) 2022 Feb 23 (pp. 7-12). IEEE.

Robotics and IoT with AI

Robotics and IoT systems involve processing of sensor data in order to perform analysis of events that lead to better decisions either by humans or autonomous agents such as mobile robots. We mainly work on real-time object detection with cameras/lidars as well as processing of other sensor data such as Inertial Motion Unit (IMU)
  • Nordt J, Ambegoda T, Chen B, inventors; Enway GmbH, assignee. Cleaning apparatus and method for operating a cleaning apparatus. United States patent application US 16/764,144. 2020 Sep 3.

Document Analysis and Digitization

The world is moving towards being paperless. This means, instead of filling out a form on paper, people are increasingly expected to fill out a digital version of it on a computer. However, to this date there are billions a paper forms that are already filled out, even if we decide to ignore the paper forms being filled even today. Digitizing the contents of those forms preserving the form's structure is critical for archiving, searching and analytics that would directly benefit decision making in business as well as government. Our efforts focus on developing methods and tools to digitize forms filled in Sinhala.