The slide deck for this talk → https://docs.google.com/presentation/...
More about the 2023 Geo for Good Summit → https://g.co/earth/geoforgood23

❤ DESCRIPTION:
If you've worked with optical remote sensing images, you've probably had to deal with clouds, cloud shadows, and other atmospheric artifacts. Flagging and masking these artifacts is a common, yet surprisingly challenging step in most image processing workflows. In this session, we introduce Cloud Score+, a general AIpowered approach for cloud and shadow detection that leverages both spatial and temporal context to produce a flexible, perpixel usability score. We'll cover the methods, then demonstrate how Cloud Score+ can be used outofthebox to generate cloudfree composites, quality mosaics, and cleaner time series. We'll also share access to a preview collection with Cloud Score+ products for the Sentinel2 archive.

#remotesensing #EarthEngine #cloudcomputing #GeographicInformationSystem #GIS #ClimateAnalysis #EarthObservation #climatechange #sustainability #geospatial #GeoAnalysis

TIMESTAMPS:
0:00 Introduction to Sentinel2
2:05 Sentinel2 Innovations
4:39 Cloud Detection Training
10:43 Video Analysis Techniques
11:56 ASIM Pretraining Explained
16:09 FineTuning with Weak Supervision
22:50 Cloud Score+ Mechanism
27:50 Cloud Score+ Masking Function
29:55 Earth Engine Integration cs vs cs_cdf
33:25 Cloud Score+ Image Collection Overview
35:00 `linkCollection` Function Introduction
37:00 Land Cover Mapping Case Study
47:14 Cloud Score+ Results
49:20 Evaluating Accuracy
50:34 Threshold Testing Strategies
53:35 Audience Q&A Session
1:01:20 Data Catalog Feedback Discussion
1:03:11 Exploring the Cloud Score+ Collection
1:10:05 Cloud Score+ Composite Tool
1:12:47 Closing Remarks and Acknowledgements

SPEAKERS:
Valerie Pasquarella, Google
Andréa Puzzi Nicolau, Spatial Informatics Group