Research data sources and models
Opened this issue · 5 comments
We are now assuming we will use Sentinel-2.
- What are the alternatives?
- Are they compatible/convertable?
- What are the differences? Resolution, frequency of updates, ease of access, cost?
Based on current information it looks like the one shared by Jacopo is great:
https://huggingface.co/spaces/ibm-nasa-geospatial/Prithvi-100M-sen1floods11-demo
Are there alternatives?
What are the pros/cons?
Some findings on Sentinel-1 models
ESRI/ArcGIS has a model for this:
- https://www.esri.com/arcgis-blog/products/arcgis/imagery/flood-inundation-mapping-using-radar-ai/
- https://developers.arcgis.com/python/samples/flood-inundation-mapping-using-sar-data-and-deep-learning/
NVIDIA did some open source work:
- https://ar5iv.labs.arxiv.org/html/2107.08369
- https://github.com/sidgan/ETCI-2021-Competition-on-Flood-Detection
Copernicus publicshed a paper as well:
Copernicus flood dataset, maybe it is this one?
https://emergency.copernicus.eu/mapping/ems/p05-modelled-flood-extent-major-events
Ok I looked a little bit more into the links I sent before, and appararently Nasa had a competition in 2021 to do exactly what we want to do :)
Here is the competition: https://nasa-impact.github.io/etci2021/
The winner was this team: https://github.com/YZArren/ETCI2021
The second place was the Nvidia repo I sent before: https://github.com/sidgan/ETCI-2021-Competition-on-Flood-Detection
They use Sentinel-1 SAR data so radar not optical. They also distinguish between water bodies and flood.
@Feynlady, could you look into this a bit more and try to get one of these models to work in Colab?
Another relevant dataset is provided in this challenge (shared by Jacopo):
https://ieee-dataport.org/competitions/2024-ieee-grss-data-fusion-contest-flood-rapid-mapping