This month we explore various data challenges, tutorials, and other resources. There are also numerous machine learning methodologies and applications, including information on applying models in disparate geophysical and geospatial datasets, building better dust detection models and classifying drivers of global forest watch, to name a few. Read September’s ML4EO market news roundup below.
Methodologies & Applications
- Learning super-resolution for Sentinel-2 images with real ground truth data from a reference satellite
- Crop map for Belgium is available, showing the main crop type on each individual field cultivated in the 2020 growing season
- Copernicus Global Land Service releases annual 100m global land cover map, covering 2015-2019 with 10 classes
- Mapping 20 years of corn and soybean across the US Midwest at 30 cm resolution using the Landsat archive
- Predicting county-scale maize yields with publicly available data
- Spatial validation reveals poor predictive performance of large-scale ecological mapping models
- Results of Uncertainty Mapping in the West Musgraves, Australia
- Building better dust detection using data received from Earth observation satellites
- Area Monitoring — Bare Soil Marker: Detecting ploughing, harvest and similar events
- Google Earth Engine cloud computing platform for remote sensing big data applications: A comprehensive review
- Machine Learning and the End of Atmospheric Corrections: A Comparison between High-Resolution Sea Surface Salinity in Coastal Areas from Top and Bottom of Atmosphere Sentinel-2 Imagery
- Classifying drivers of global forest loss
- Graph Convolutional Networks for hyperspectral image classification (access the code on GitHub)
- More Diverse Means Better: Multimodal deep learning meets remote-sensing imagery classification
- Monitoring sustainable development by means of earth observation data and machine learning: A review
- AiDash launched a predictive vegetation management product for utilities that leverages the UP42 Earth observation marketplace
Data Access & Validation
- SpaceNet 6: Expanded dataset release
- Cloud-Spotting at a million pixels an hour: How I learned to draw clouds in satellite imagery during a data labeling competition
- Announcing the winners of the data labeling contest
- Planet, KSAT, and Airbus awarded the first-ever global contract to combat deforestation. They will provide universal access to high-resolution satellite monitoring of the tropics
- Narrowband, Part Two: An exploratory look into the inner workings of GroundWork, a web-based segmentation labeling tool
Standards Research & Innovation
- Presentation Recordings: Cloud Native Geospatial Outreach Day 2020
- Harmonized Landsat Sentinel-2 project provides Analysis Ready Data to increase the capability of analysis
- Analysis Ready Sentinel-1 Backscatter Imagery
Tutorials, Webinars (recorded) & Resources
- Browse NASA Earth Science stories by location
- Scale-up your EO-learn workflow using Batch Processing API
- PODCAST – Eyes on Earth: Open Training Data with Anne Hale Miglarese
- PODCAST – Scene From Above: Platform Proliferation, including discussions on STAC
- TUTORIAL – How to use GroundWork and Radiant MLHub to create and publish open training data
- TUTORIAL – How to combine global climate models and land cover data to estimate the regional effects of climate change years into the future
- TUTORIAL – How to implement augmentations for multispectral satellite images segmentation using Fastai-v2 and Albumentations
- TUTORIAL – Deploying the SpaceNet 7 baseline on AWS
- TUTORIAL – Exploratory Data Analysis on satellite imagery using EarthPy
- OPINION – Machine learning should combat climate change
- 4 must-have JupyterLab Extensions for geospatial data science
- Call for papers (special issue): “How the Combination of Satellite Remote Sensing with Artificial Intelligence Can Solve Coastal Issues” – manuscript deadlines December 31, 2020
Data Challenges & Conferences
- SpaceNet 7: Multi-Temporal Urban Development Challenge
- DengAI: Predicting Disease Spread
- Circle Finder Prize Competition to segment satellite imagery to detect, delineate, and describe circular-shaped features
- Leveraging Machine Learning to Minimize Climate Change Impacts: To predict how Nature-Based Solutions help societies and ecosystems adapt to climate impacts
- AI for Earth Sciences: NeurIPS 2020 Workshop, Saturday December 12, 2020 – Papers due October 7, 2020
- Tackling Climate Change with Machine Learning: NeurIPS 2020 Workshop, Friday or Saturday December 11/12, 2020 – Papers due October 6, 2020
What are we missing? Contact Louisa@radiant.earth