Dineth Perera

Electrical & Electronic Engineering Undergraduate

Research interests in computer vision, remote sensing, signal processing, generative AI, and medical imaging.

University of PeradeniyaGPA: 3.70/4.00Sri LankaOpen to research internships
About

About

I am a third-year Electrical and Electronic Engineering undergraduate at the University of Peradeniya, Sri Lanka. My current work focuses on remote-sensing computer vision, especially binary change detection, semantic segmentation, visual state-space models, and dataset preparation for disaster and wildfire analysis. I am interested in research internships in computer vision, remote sensing, signal processing, generative AI, and medical imaging.

Research

Research & Current Work

Current work across remote-sensing change detection, dataset preparation, and signal-processing-based sensing.

Change Detection

Ongoing research · Remote-sensing change detection

An ongoing binary change detection project using MambaVision-based bi-temporal feature extraction and region-boundary refinement for remote-sensing imagery.

Ongoing research
  • Bi-temporal pre/post image input.
  • Shared encoder design.
  • Temporal difference features.
  • Region and boundary-aware refinement.
  • Evaluation on standard binary change detection datasets.
Change DetectionMambaVisionRemote SensingBoundary Refinement

WildFire-S2

Dataset preparation · Wildfire remote sensing

A Sentinel-2 bi-temporal dataset preparation project for wildfire burned-area change detection using pre-fire images, post-fire images, and binary burned-area masks.

Dataset preparation / documentation
  • Sentinel-2 imagery.
  • Pre-fire / post-fire image pairs.
  • Binary burned-area masks.
  • Dataset organized for supervised change detection.
  • Includes dataset documentation and loading examples.
WildfireSentinel-2Dataset PreparationChange Detection

DiTwa Disaster Tri-Temporal Dataset

Dataset preparation · Disaster remote sensing

An ongoing dataset preparation project for tri-temporal disaster analysis using imagery from multiple stages of a disaster event.

Ongoing dataset preparation
  • T1: pre-disaster image.
  • T2: post-disaster image.
  • T3: on-disaster image.
  • Intended for multi-temporal disaster change analysis.
  • Metadata and dataset structure under preparation.
Disaster Remote SensingTri-temporal DataDataset Preparation
Publications

Publications & Manuscripts

Accepted and under-review research outputs.

A Controlled Benchmark of Visual State-Space Backbones with Domain-Shift and Boundary Analysis for Remote-Sensing Segmentation

  • Accepted, IGARSS 2026
  • Nichula Wasalathilaka, Dineth Perera, Oshadha Samarakoon, Buddhi Wijenayake, Roshan Godaliyadda, Vijitha Herath, Parakrama Ekanayake
Accepted
ContributionContributed to experiments, benchmarking, analysis, and manuscript preparation.
SummaryControlled comparison of visual state-space backbones and baseline models for remote-sensing semantic segmentation.
Remote SensingSemantic SegmentationVisual SSMs

MambaRefine-CD: MambaVision with Region-Boundary Temporal Refinement for Remote-Sensing Change Detection

  • Manuscript / Under Review, MERCon 2026
  • Dineth Perera, Thaariq Firdous, Oshadha Samarakoon, Buddhi Wijenayake, Roshan Godaliyadda, Parakrama Ekanayake, Vijitha Herath
Under Review
ContributionDeveloping the change detection architecture, experimental setup, and analysis.
SummaryOngoing manuscript on MambaVision-based binary remote-sensing change detection.
Change DetectionMambaVisionRemote Sensing
Selected Projects

Selected Projects

Research projects are listed above. This section includes other technical work kept in compact form.

Research projects are listed above.

Signal Processing / Machine LearningProject2025

Real-Time Elephant Detection Using Infrasound Analysis and Machine Learning

Signal-processing and machine-learning project exploring low-frequency infrasound features for elephant detection.

InfrasoundSignal ProcessingMachine Learning
Signal and Image ProcessingExperiments

Wavelets / Contourlets / XLET-NSST Experiments

Experiments with multiscale signal and image transforms, including wavelet, contourlet, XLET, and NSST workflows.

Signal ProcessingImage ProcessingMATLAB
Remote-Sensing SegmentationResearch code

Controlled Visual SSM Benchmark Implementation

Implementation work connected to controlled visual SSM benchmarking for remote-sensing semantic segmentation.

PyTorchRemote SensingSegmentation
Skills

Skills

Grouped technical skills and research tools.

Programming

  • Python
  • C/C++
  • MATLAB
  • LaTeX
  • Git
  • Linux

Machine Learning / Computer Vision

  • PyTorch
  • OpenCV
  • TensorFlow
  • CNNs
  • Transformers
  • visual state-space models
  • segmentation
  • change detection

Remote Sensing

  • Sentinel-2
  • semantic segmentation
  • change detection
  • dataset preparation
  • boundary-aware evaluation

Signal Processing

  • FFT/PSD analysis
  • filtering
  • spectral features
  • vibration analysis
  • infrasound analysis

Tools

  • GitHub
  • Hugging Face
  • Google Scholar
  • arXiv
  • LaTeX
Awards

Awards

Selected academic prizes, examination results, and competition outcomes.

  • 2026
    IGARSS 2026 paper accepted
    Controlled visual SSM benchmark for remote-sensing segmentation.
  • 2025
    SLTTECHNOVATION 2025 - Semifinalist
    Smart Waste Water Management solution.
  • 2024
    Haxtreme 4.0 - 7th Place
    Team Binary Brains.
  • 2024
    SLIIT Extreme 4.0 - Finalist
    Team Binary Brains.
  • 2024
    UOJ Coder v3.0 - Finalist
    Programming competition finalist.
  • 2021
    P.D.S. Kularathne Memorial Prize and N.B.M. Mediweka Memorial Prize
    Best student selected to the Faculty of Engineering.
  • 2021
    S.J. Munasinghe Memorial Prize
    Highest Z-score at the G.C.E. Advanced Level Examination.
  • 2021
    G.C.E. Advanced Level Physical Science Stream
    National Rank 68, District Rank 4.
CV & Links

CV & Research Links

For research internship opportunities and collaborations, please refer to my CV and research profiles.

Contact

Contact

Direct contact information and research profiles.

Research conversations, applications, and collaborations

Email is the most direct contact channel for research internships, lab applications, and collaborations.