Webinar - Unlocking Machine Learning Potential Using openEO скачать в хорошем качестве

Webinar - Unlocking Machine Learning Potential Using openEO

The openEO API instance offered in the Copernicus Data Space Ecosystem enables the creation of large-scale maps using time series pixel classification at relatively low operational costs. In addition, workflows remain simple and maintainable as all complexity is handled by the backend. In this webinar held on 18 February 2025, Jeroen Dries, the product lead of openEO at VITO, shows how several mapping projects have already successfully used this technology and even managed to integrate base models that allow them to create classification pipelines that become more and more generalised without sacrificing accuracy. See the following timestamps for easier navigation: 0:00 Introduction 1:02 OpenEO feature: ML capabilities at scale 4:21 OpenEO: delivering EO & Machine Learning 'as a service' 6:53 WorldCereal workflow 9:18 Deep Dive: Dynamic landcover mapping CDSE Documentation ➡️ https://documentation.dataspace.coper... Jupyter Notebook ➡️ https://github.com/Open-EO/openeo-com... 31:00 Use of ONNX & UDF https://github.com/Open-EO/openeo-com... 35:00 Q&A 35:10 What is the advantage of using OpenEO over Planetary Computer by Microsoft? 37:05 Can we do cloud masking using this method provided in GEE API: ee.ImageCollection('COPERNICUS/S2_CLOUD_PROBABILITY') 37:53 How do you treat cloud coverage on areas with persisting clouds? 38:56 What does the "ellipsoidal" stand for in "sigma0-ellipsoidal" in S1 data inputs? 39:28 Should we run this computation on Jupyter, or can we run it on the open EO directly? 40:19 Is it possible to compute vegetation indices and use them as co-predictors in the classification or regression models? 41:01 How many hyper parameters have been considered during the classification? 41:29 Do you have an example of how I can fine-tune an existing model (e.g. the Presto) using openEO? 43:13 Is VITO involved in producing ESA CCI LULC 300m. It seems that it is only available up to 2022. Can we use this pipeline to produce 2023 and 2024 LULC? Is the model used like ESA CCI LULC? 44:16 Can one deduce NDVI using the tool? 44:42 What is a difference between GEE and Microsoft Planetary computer and OpenEO? 47:00 Do you support other data (Landsat-*, Harmonized Landsat Sentinel ...) or only the Copernicus data ? 47:45 Are the statistics (median, StDev, quartiles) calculated per pixel or segment from the image composites over the observed period? 48:04 Conclusion Useful links: OpenEO Documentation ➡️ https://documentation.dataspace.coper... CDSE Forum ➡️ https://forum.dataspace.copernicus.eu/ Gallery ➡️ https://dataspace.copernicus.eu/gallery Newsletter ➡️ https://dataspace.copernicus.eu/subsc...