my photo

Nikolaos Anastasiou

Surveying and Geoinfomatics Engineer

I received my diploma in Rural, Surveying and Geoinformatics Engineering from the National Technical University of Athens in 2025. My diploma thesis focused on the development and evaluation of deep learning models for wildfire spread prediction using remote sensing data. I completed a three-month internship at the National Observatory of Athens, focusing on Earth observation and wildfire mapping. I am currently working at the Athena Research Center, where I am involved in training foundation models for fire detection from satellite data. My primary research interests include remote sensing and deep learning, with an emphasis on wildfire-related applications.

📧 nikolas619065@gmail.com   |   📞 (+30) 6945692088   |   📃 Download CV

Experience

Athena Research Center

Collaboration with the PREFERRED project team at ARCHIMEDES Unit • April , 2025 — Present

  • Comparison and fine-tuning of SatMAE-type foundation models pre-trained on multispectral data for wildfire detection from Seviri satellite data, and comparison with Convolutional Neural Networks
  • Implementation of a Masked Autoencoder model with a Vision Transformer backbone in Python using the PyTorch framework, and pre-training with thermal satellite data

National Observatory of Athens

Internship at unit Beyond Earth Observation Center • July, 2024 — Sept, 2024

  • Mapping of wildfires using Sentinel-2 data in QGIS software
  • Automation of the mapping process utilizing the frameworks OpenDataCube, GeoPandas, Rasterio, and Xarray in Python
  • Automation of the data extraction process from daily fire department reports (from .pdf files to a database) to automatically calculate statistics
  • Development of tools in Python to calculate statistics from burned areas, such as land use coverage percentages, NATURA2000 areas, previous burned areas, etc.
  • Creation of a Machine Learning Dataset for fire risk estimation
  • Participation in two Copernicus Emergency Management Service (EMS) projects

Education

National Technical University of Athens

MEng, School of Rural, Surveying and Geoinfomatics Engineering • 2020 — 2025

  • Dioploma Thesis

As part of this thesis, the models U-Net2D, U-Net3D, and Vision Transformer were trained using remote sensing data with the aim of predicting the final burned area from wildfires. The models were implemented using the PyTorch framework in a Python environment. The dataset included meteorological variables, NDVI and LAI indices, previously burned areas, comprising a total of 28 variables. The models were trained and making predictions using data from 5 days before and 5 days after the fire events, given an ignition point. Additionally, data were collected to analyze how forest fires spread and which variables contribute to their propagation, using Explainable AI techniques. Thesis PDF, Slides, data and the developed models are open source and available on:
Thesis PDF, Thesis Slides, Thesis GitHub

Thesis Supervisor: Dr. Ioannis Papoutsis

Projects

Developed in Python using frameworks Xarray, GeoPandas, odc, Rasterio, Shapely, etc., which, by providing an area with 4 coordinates and the start date of a wildfire, performs fully automated mapping using Sentinel-2 data. The output includes a tiff file with the DNBR index, a shapefile with the wildfire polygon (geometrically corrected), and a shapefile with the classification of the burned area into 4 categories. Detailed information is available on GitHub

Developed in Python, which, by providing a shapefile of the burned area polygon and shapefiles such as NATURA2000, Corine Land Cover, etc., calculates the percentage of each category within the wildfire polygon. In other words, it calculates what percentage of a burned area is NATURA2000 or any other category. Detailed information is available on GitHub

A Python script for solving 2D geodetic networks with complex geometry, using as input only the network points. The script solves the network and returns a PDF file with the network’s geometric elements and solution param- eters, as well as a plot of the network’s shape. It is very useful for identifying if a specific point causes poor geometry in the network, thus affecting its accuracy

A Python script for solving a 1D elevation network with input data consisting only of orthometric heights. The script returns a PDF file with the network’s solution parameters and the corrected heights. It is very useful for identifying problematic heights in the network

Skills

Languages

Python, C++

Frameworks

Xarray, GeoPandas, Pandas, NumPy, Matplotlib, Pytorch

Programms

QGIS, ArcGIS, AutoCAD(2D), Git, Latex, VSCode, PyCharm, JupyterNotebook

Publications

Nikolaos Anastasiou, Spyros Kondylatos and Ioannis Papoutsis, "Wildfire spread forecasting with Deep Learning"

Submitted at IEEE Access under review, ArXiv Submission: https://arxiv.org/abs/2505.17556

Distinctions

Professor Χ. Papakyriakopoulou award for the academic year 2021-2022

National Technical University of Athens • 2023

The award is granted by the School of Applied Mathematical and Physics Science to students who have the highest average grade in the mathematics courses of the academic year, in each school of the National Technical University of Athens (NTUA).

Social Links

  • Github: https://github.com/nikos230
  • LinkedIn: https://www.linkedin.com/in/nikos-anastasiou/
  • Website: https://nikos230.github.io