Trillium Australia - 2021 in retrospect
Well, what can I say? It has been a wild year for Trillium in Australia and worldwide. Our amazing researchers and partners dug deep during difficult times to produce superb research outcomes and insights for a host of projects. My heartfelt thanks to you all - you have been an inspiration!
We broke the ice on 2021 with our ML for Climate Change (ML4CC) project, delivering a MLOPs toolkit for flood extent modelling, fully integrated with Google Cloud and accessible by ML novices.
February saw us working with the ACT Rural Fire Service, AFAC and other key bushfire organisations to broaden our understanding of bushfire management systems. Building on an initial Concurrent Design Workshop at UNSW ANCDF, we summarised how bushfires are fought in Australia and we mapped out the opportunities for technology to help manage increasingly large fires. Our findings are released in an open report.
In May we worked with partners D-Orbit, Unibap, ESA Phi-Lab and the University of Oxford to build our first machine learning payload. Part of our Networked Intelligence in Orbit (NIO) vision, the WorldFloods ML Payload was launched successfully aboard the D-Orbit Wild Ride mission in late July and was hosted on the Unibap Nebula SpaceCloud computing module. We demonstrated rapid processing of large Sentinel-2 data cubes to small and easily-downloaded vector flooding maps for use on the ground. In a world-first, we also showed how a neural network could be re-trained to successfully process data from a completely new instrument, without changing any of the flight qualified code.
Our second Networkd Intelligence in Orbit (NIO) project is due for launch on January 10th 2022, on SpaceX Transporter-3. The ML Payload links to a novel hyperspectral imager made by VTT and we are really excited about the broad applications of this technology - land use mapping, vegetation types, burn scars, tracking active fires, emissions compliance, and many more.
Trillium Australia also started a new project in lead up to COP26 to understand how machine learning techniques could help measure Blue Carbon in ocean systems, leading to meaningful interventions in the future. We expect this work will lead to new interdisciplinary research challenges in 2022.
Finally, Trillium USA and Europe ran super-successful Frontier Development Lab research programs in 2021, leading to some amazing results. You can check these out at https://frontierdevelopmentlab.org/publications, or go to https://spaceml.org/ for code and data.
I’m keenly looking forward to 2022, when Trillium Australia will be hosting FDL Bushfires, continuing our work on distributed intelligence in orbit (and on Earth) and helping humans be better stewards of our planet.
Have a safe and wonderful holiday, and an even better New Year.
Ad astra per algorithmos.
Cormac