Dong Pham
About Me
I am a researcher and PhD specializing in continental-scale land-cover mapping and deep learning for Earth observation. At the EO Lab, University of Greifswald, I work on scalable and consistent land-cover characterization across extensive spatial and temporal domains. My work integrates research and engineering by developing transferable geospatial models and datasets and deploying them on high-performance computing (HPC) and cloud infrastructures.
Research Interests
- Large-scale remote sensing application
- Methodological and algorithmic development
- Machine learning for Earth observation
- Time-series analysis
- Model transferability and domain adaptation
- High-performance and distributed computing
Highlights
I am the creator of:
| Method | Description | Publication | Application |
|---|---|---|---|
| Multi-classes Neural Network regression | This method quantifies land cover fractions using Neural Network and regression-based unmixing. | DOI | 1, 2 |
| Temporal Encoding and deep learning augmentations | This method encodes raw time-series, combined with augmetation methods to enhance land cover mapping's transferability. Read more in this blog. | DOI | 1 |
| Super-resolution and center-patch classification | The super-resolution methods enhances the spatial resolution of historical Landsat data to 10-m. The center-patch classification learns the contexts of the surrounding to classify the center pixel. Read more in this blog. | DOI | 1 |