Skip to content
View devonrama's full-sized avatar
🚧
🚧

Block or report devonrama

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
devonrama/README.md

Devon Rama Wikunanda

Undergraduate Geodetic Engineering student at Universitas Diponegoro, Semarang, Indonesia. I build remote-sensing and geospatial ML pipelines for environmental monitoring SAR optical change detection, InSAR deformation, and land cover classification with an emphasis on validation rigour and honest scoping.

Focus Areas

  • SAR–optical fusion (Sentinel-1 + Sentinel-2) for change detection and land cover mapping
  • InSAR time series for land subsidence and coastal deformation (LiCSBAS)
  • Geospatial machine learning (XGBoost, LightGBM, Random Forest) with spatial cross validation
  • Google Earth Engine + Python pipelines (rasterio, geopandas, GDAL) and spatial databases (PostGIS)

Featured Projects

Sentinel-1 / Sentinel-2 fusion pipeline screening a ~38,800 ha Indonesian oil palm concession for forest loss after the EU Deforestation Regulation cutoff (2020-12-31). The primary detector is unsupervised SAR change point detection radar penetrates the near permanent tropical cloud cover that blinds optical only methods confirmed by Sentinel-2 optical drops. Detected loss is cross validated against three independent satellite references (RADD, Hansen GFC, JRC TMF), reaching ~0.81–0.85 strict-pixel and ~0.87–0.90 patch area precision. A supervised XGBoost model is included as a documented comparison, with its honest non-result reported as part of the contribution. Capability demonstration, not a legal grade compliance product.

Stack: Google Earth Engine, Python, rasterio, XGBoost, Folium

Sentinel-1 SAR flood frequency and LiCSBAS InSAR subsidence mapping for the Semarang–Demak coast (2020–2025), fused into a per-kelurahan coastal-flood vulnerability typology. Built for the IEEE GRSS REACT EO4SDG challenge. SDG 11.5 / 13.1.

Stack: Sentinel-1, LiCSBAS, ESA SNAP, Python, geopandas

End-to-end remote sensing pipeline mapping industrial land cover in the Karawang corridor, West Java, via Sentinel-1 / Sentinel-2 fusion and XGBoost classification. Produces 1,063 industrial polygons with documented human-in-the-loop refinement and an explicit limitations section.

Stack: Python, rasterio, geopandas, XGBoost, ESA SNAP

Drone imagery photogrammetry with GCP-based geodetic validation orthophoto and DSM generation and accuracy assessment using OpenDroneMap / WebODM.

Stack: OpenDroneMap, WebODM, GCP survey, Python

Async REST orchestrator for resource constrained binary tools, demonstrated with COLMAP for Structure-from-Motion reconstruction. Single tenant FastAPI service with subprocess timeout, payload validation, and VRAM bounded preprocessing.

Stack: Python, FastAPI, COLMAP

Technical Background

Domain Tools and Methods
Remote sensing Sentinel-1/2, Landsat, ESA SNAP, LiCSBAS, Google Earth Engine
Change detection / InSAR SAR backscatter time-series, change-point detection, InSAR deformation
Geospatial ML XGBoost, LightGBM, Random Forest, scikit-learn, spatial cross-validation
Geospatial Python rasterio, geopandas, GDAL, shapely
Databases PostgreSQL, PostGIS
Backend FastAPI, async subprocess orchestration
Geodetic computation Map projections, least-squares adjustment, satellite geodesy

Currently Learning

  • Deep learning for semantic segmentation in remote sensing
  • Operational SAR/InSAR monitoring workflows (LiCSBAS, time-series)
  • Reproducible geospatial ML pipelines determinism, spatial CV, validation rigour

Contact

Pinned Loading

  1. eudr-deforestation-demo eudr-deforestation-demo Public

    SAR–optical (Sentinel-1/2) change-detection demo screening an Indonesian oil palm concession for EUDR post-2020 deforestation. Capability demonstration, not legal-grade.

    HTML

  2. coastal-subsidence-semarang-demak coastal-subsidence-semarang-demak Public

    Sentinel-1 SAR flood-frequency and LiCSBAS InSAR subsidence mapping for the Semarang–Demak coast (2020–2025), fused into a per-kelurahan coastal-flood vulnerability typology. SDG 11.5 / 13.1.

    Python 2

  3. karawang-industrial-mapping karawang-industrial-mapping Public

    End-to-end geospatial ML pipeline for industrial zone mapping in Karawang, Indonesia using Sentinel-1/2 multi-sensor fusion

    Python 2

  4. uav-photogrammetry-accuracy-study uav-photogrammetry-accuracy-study Public

    UAV photogrammetry with GCP-based geodetic validation. Demonstrates orthophoto and DSM generation from drone imagery using OpenDroneMap.

    3

  5. Headless-SfM-Engine Headless-SfM-Engine Public

    Async REST orchestrator for resource-constrained binary tools, demonstrated with COLMAP for SfM reconstruction. Single-tenant FastAPI service with subprocess timeout, payload validation, and VRAM-b…

    Python 1

  6. devonrama devonrama Public

    3