Presto: Lightweight, Pretrained Transformers for Remote Sensing Timeseries | Gabriel Tseng | Harvest
Gabriel Tseng presented at the SERVIR's Geo-AI Working Group on May 1, 2024.
Gabriel Tseng is a PhD student at McGill University / Mila and a member of NASA Harvest’s Machine Learning team. His research focuses on how we can use remote sensing and machine learning to better understand the planet, with a focus on agriculture. His work spans dataset creation, methodological research and the deployment of models alongside end users.
In this talk, Gabi presented Presto, a lightweight pre-trained foundation model for remote sensing data. Presto is designed to make maps at scale, even when there are few labels. Gabi discussed how his experience making maps influenced the construction of Presto, and how it aims to meet the needs of remote sensing practitioners.
Find Gabi's Presentation here: https://docs.google.com/presentation/...
Learn more about Presto and related work here:
https://github.com/nasaharvest/presto
https://github.com/WorldCereal/presto...
https://arxiv.org/abs/2304.14065
=======================================
Website: https://tinyurl.com/servir-geo-ai-wg
Geo-AI WG Google Group: https://tinyurl.com/join-geo-ai-wg
Recurring meeting invite
SERVIR's Geo-AI Working Group Bi-Weekly Meeting
Bi-weekly Wednesday, 10:00 – 11:00 AM CT
Google Meet joining info
Video call link: https://meet.google.com/jmo-ojko-imx
Add invite to your Google Calendar:
1) Go to the Official TFWG Calendar: https://calendar.google.com/calendar/...
2) Click "+ Google Calendar" in the very bottom right corner to add to your Google Calendar.
#geospatial #machinelearning #deeplearning #foundationalmodel #transformers #agriculture #landcover #ecosystem #mapping #nasa #servir #thegeoict #geoai #mcgilluniversity #nasaharvest