Happy New Year from the ML4EO Market News team! Highlights in December include the release of Landsat Collection 2 in COG format, new training datasets, tutorials, including how to build a model for Radiant Earth’s latest data competition, and various papers focusing on impactful uses of machine learning in reducing and responding to climate change.
Methodologies & Applications
- Generating Synthetic Multispectral Satellite Imagery from Sentinel-2 (watch the recording of the presentation here)
- Semantic Segmentation of Medium-Resolution Satellite Imagery using Conditional Generative Adversarial Networks (watch the recording of the presentation here)
- A satellite-based spatio-temporal machine learning model to reconstruct daily PM2.5 concentrations across Great Britain
- Fusing Optical and SAR time series for LAI gap filling with multi-output Gaussian processes
- LandCoverNet: A global benchmark land cover classification training dataset (watch the recording of the presentation here)
- Real-time flooding alerts for drivers? ODU researchers are using machine learning to make it happen
- Artificial Intelligence, Machine Learning, and Modeling for Understanding the Oceans and Climate Change
- A clever strategy to deliver COVID aid—with satellite data. Togo used image analysis algorithms to target economic support for the most vulnerable
- Using very‐high‐resolution satellite imagery and deep learning to detect and count African elephants in heterogeneous landscapes
- Benjamin Akera: Machine Learning for Glacier Monitoring in the Hindu Kush Himalaya
- Monitoring the Impact of Wildfires on Tree Species with Deep Learning
- ForestNet: Classifying Drivers of Deforestation in Indonesia using Deep Learning on Satellite Imagery
- Predicting the Solar Potential of Rooftops using Image Segmentation and Structured Data
Data Access & Validation
- Semantic Segmentation of Crop Type in South Sudan released on Radiant MLHub
- Semantic Segmentation of Crop Type in Ghana released on Radiant MLHub
- Open Cities AI Challenge Dataset released on Radiant MLHub
- iSAID: A large-scale dataset for instance segmentation in aerial images
- First images from ocean-monitoring satellite released. The Copernicus Sentinel-6 Michael Freilich satellite delivers promising first altimeter data which can be used to map Earth’s rising sea levels
- Four insights from Azavea to help you manage data labeling for machine learning projects
- USGS releases the most advanced Landsat archive to date – Landsat Collection 2 is distributed in Cloud Optimized GeoTIFF format and includes a STAC catalog
- Capella unveils the world’s highest-resolution commercial SAR imagery
- Zindi and SAEON fill in the air quality data gaps for African cities
- Copernicus Digital Elevation model provided as Cloud Optimized GeoTIFFs
Standards Research & Innovation
- The first STAC API 1.0 release: 1.0.0-beta.1
- Geemap now supports loading publicly hosted Cloud Optimized GeoTIFF and STAC with only one line of code
- GDAL 3.2.1 is released
- SpaceNet 7 Results: ‘Tis the Season – This blog details the innovations that vastly improved upon the baseline model
Tutorials, Webinars (recorded) & Resources
- TUTORIALS
- Benchmark – wind-dependent variables: Predict wind speeds of tropical storms
- STAC, COG, Python, and QGIS – A simple guide to search for satellite data and visualize them in QGIS and Python
- Change Detection with LandTrendr – Extract long-term change information using the LandTrendr algorithm, available in ArcGIS Pro 2.7 with the Image Analyst extension.
- Detecting Changes in Sentinel-1 Imagery (Part 3)
- WEBINARS
- Why satellite imagery, AIS, and machine learning are making waves in maritime surveillance hosted by UP24
- Machine learning and the satellite revolution for climate resilience – speakers from Cloud2Street, Radiant Earth, and NASA SERVIR
- GeoAI: Applying Deep Learning to Geospatial Data organized by IEEE GRSS and ISPRS
- PODCASTS
- Yale Climate Connections with ClimateTRACE’s Gavin McCormick – Pinpointing sources of carbon pollution using satellite data, AI, and machine learning to monitor global emissions in real-time
- The Scene from Above – GEE with Keiko Nomura
- ML Minutes with Jimi Crawford – How can satellite imagery enable sustainable supply chains?
- RESOURCES
- The 5 pitfalls to avoid when accessing and analyzing satellite imagery
- Explore open spotlight talks and poster presentations of two NeurIPS 2020 Workshops: Tackling Climate Change with Machine Learning and AI For Earth Sciences
- Machine learning and object detection in spatial analysis
- OPINIONS
Data Challenges
- Wind-dependent Variables: Predict Wind Speeds of Tropical Storms
- CGIAR Crop Yield Prediction Challenge
Conferences & Workshops
- The Climate Crisis AI Hackathon (Jan 22-24, 2021)
- The UofT AI Conference (Jan 15-16)
What are we missing? Contact Louisa@radiant.earth