crisjosil Goto Github PK

👋 Hi, I’m Cristian, I’m interested in extracting information from satellite imagery and geospatial data using statistical and machine learning algorithms. Currently I am a postdoctoral researcher developing novel methods for near-real time monitoring of climate extremes and carbon landscapes from SAR imagery.

Some of my repos include:

• Our work covering the Kakhovka Dam explosion in Ukraine here. I trained and tested a few different architectures of deep learning models on a dataset that I curated covering several flood events.
• A post about how to use the Google Earth Engine python API directly in QGIS. In this case, I show an example of flood monitoring here.
• A side project working on image object detection with the YoloV8 model pretrained by ultralytics here. I also containerised this as an app and deployed it using microsoft azure. You can interact with this app here (may take some time to load as it is the free version without pre-warmed workers). The plan is to retrain this model for aerial object detection (in process).
• After hearing in the news about a wildfire that could become the largest in the UK, I checked for satellite imagery covering the area and used a well known method for mapping fires. Check some outcomes of this here
• PolSAR paper: Repo of a peer reviewd publication using Quad-PolSAR and machine learning for crop monitoring here.
• Hydrological data API: A notebook for quering an API to fetch hydrological data from a network of IoT sensors distributed accross Scotland. I also use folium and geopandas for data analysis here.
• A colab notebook which downloads surface soil moisture data from the ECMWF-ERA5 dataset via the Google Earth engine python API. It reads the areas of interest from a polygon, filters the image collection and downloads the desired data (.csv)here
• I have several private repos including (and others):
• A Dockerised ETL pipeline for ingestion of IoT data into an Azure hosted PostgreSQL DB. This process is triggered by event-driven azure functions (serverless computing).
• A Dockerised ETL pipeline for download, pre-processing and serving satellite imagery.