Cloud Score+: A Clear View for Geospatial Data (using weekly supervised video analysis)| Geo4Good'23

👉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 AI-powered approach for cloud and shadow detection that leverages both spatial and temporal context to produce a flexible, per-pixel usability score. We'll cover the methods, then demonstrate how Cloud Score+ can be used out-of-the-box to generate cloud-free composites, quality mosaics, and cleaner time series. We'll also share access to a preview collection with Cloud Score+ products for the Sentinel-2 archive.

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

💻 TIMESTAMPS:
0:00 - Introduction to Sentinel-2
2:05 - Sentinel-2 Innovations
4:39 - Cloud Detection Training
10:43 - Video Analysis Techniques
11:56 - ASIM Pretraining Explained
16:09 - Fine-Tuning 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