Lakmal

N. Lakmal Deshapriya


Bio: AI/Deep Learning researcher. Always fascinated about, how human brain / intelligence works. And big believer of engineering approach of solving human intelligence by building neural nets.

Work: Researcher at Geoinformatics Center of Asian Institute of Technology (AIT), Thailand, engaging projects related to geospatial-data-analysis. These days, we are trying to bring more neural nets into our day-to-day work.

Education: M.Eng.(Computer Science), specialized in Data Science and Artificial Intelligence. I'm a student of Prof. Mathew N. Dailey. My theses (Deep Instance Segmentation and Polygonization) is about developing simple and intuitive new deep neural network architecture for Instance Segmentation.

Email: lakmal[at]ait.ac.th

GitHub: Click here !



Selected Publications:

Some of my selected publications.

Vec2Instance: Parameterization for Deep Instance Segmentation

This arXiv paper is formed from my Master's thesis work.

Link to arXiv paper: Click here !

Code of this work is in GitHub: Click here !

+


Developing methodology to map tree canopy in urban areas from low cost commercials UAVs

Link to the paper: Click here !


Landslide susceptibility mapping using logistic regression model (a case study in Badulla District, Sri Lanka)

Link to the paper: Click here !


Digital Image Processing Technique for Particle Size/Shape and Mineralogical and Textural Analysis

MUVA Siriwardana, NL Deshapriya and R Premasiri, (2013). Digital Image Processing Technique for Particle Size/Shape and Mineralogical and Textural Analysis. In 7th National conference on Earth Resource Management. University of Moratuwa, 28th June 2013. Moratuwa: Dept. of Earth Resource Engineering.

-

Teaching:

Some of short-term training courses that I've developed for project related activities at GIC-AIT.

Deep Learning for Satellite Image Analysis (Remote Sensing)

Link to Training Materials (GitHub): Click here !


Python for Geospatial Data Analysis

Link to Training Materials (GitHub): Click here !


Basics of Google Earth Engine

Link to Training Materials (GitHub): Click here !

+


Agriculture Mapping and Monitoring with Satellite Images

Link to Training Materials: Update soon

-

Main Projects:

Some of projects that I've contributed significantly so far.

Development of PyAEZ

Project Description: About developing Python package (PyAEZ) for FAO's AEZ framework. It's consisted of many mathematical crop-related models to find crop suitability under particular climate, terrain, soil. It's published in GitHub (Click here !).
My contribution: Design PyAEZ package structure. Python coding for PyAEZ package with the project team.
Client: The Food and Agriculture Organization of the United Nations (FAO). Project Period: 2019-2020.

+


Disaster Risk Assessment of Uttarakhand, India

Project Description: Disaster risk assessment project for earthquakes, floods, landslides, flash floods, and industrial hazards. And collaborative project between DHI, ERN, GIC-AIT.
My contribution: Model the risk for landslides with maximum entropy method (Maxent software). Develop web based crowd-sourcing application for settlement mapping from satellite images.
Client: World Bank. Project Period: 2016-2018.


Capacity Building on Geo-portals in Pacific Island Countries (PICs)

Project Description: Assist PICs to establish geo-portal to manage and distribute disaster related data.
My contribution: Visited 4 PICs (Tonga, Fiji, Solomon Islands and Micronesia) assisting them to establish open-source geo-portal call GeoNode, like Fiji DRM Geoportal.
Client: United Nations ESCAP. Project Period: 2016-2018.


International Disaster Charter (IDC)

Project Description: IDC is a humanitarian charter connecting space agencies and data analysis agencies, allowing government disaster agencies to obtain satellite data and value-added projects in natural disaster situations. GIC-AIT is a part of this effort, providing data analysis services.
My contribution: Project manager for 2 IDC activations (IDC-570 and IDC-575). And assisted with data analysis.
Client: International Disaster Charter (IDC). Project Period: 2016.


Provision of Satellite based Indexes for Lower Mekong Basin

Project Description: The main objective is to provide satellite based vegetation, water and drought related indexes from 2013 to 2015 for Lower Mekong Basin (covering parts of Lao, Cambodia, Thailand and Vietnam).
My contribution: Develop Shell and Python scripts to process and calculate satellite based indexes automatically.
Client: Mekong River Commission (MRC). Project Period: 2014-2015


JAXA Mini Project

Project Description: Project dedicated to develop capacity in Asia regarding application of satellite image analysis (remote sensing) in agriculture monitoring, disaster management and environment.
My contribution: Developed case studies for agriculture mapping from satellite images (Bangladesh and Cambodia) and landslide susceptibility mapping (Indonesia and Sri Lanka). And conduct training programs.
Client: Japan Aerospace Exploration Agency (JAXA). Project Period: 2013-2015.

-

Hobby Projects:

Some of my side projects that I'm working on in my free time.

Vec2Instance version 2: Vec2Instance with Attention

Link to Repository (GitHub): ...


SimpleVAE - Simple Way to Generate Smooth Latent Space

Link to Repository (GitHub): Click here !


Bio-hacking: Protein Structure Modeling with Deep Learning

Link to Repository (GitHub): Click here !


Copyright © 2020-2022 Lakmal Deshapriya.