BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Barlow Twins Earth Foundation Model - Frank Feng\, University of C
 ambridge
DTSTART:20250207T130000Z
DTEND:20250207T135500Z
UID:TALK227335@talks.cam.ac.uk
CONTACT:114742
DESCRIPTION:*Abstract*\n\nSatellite imagery provides a critical lens for m
 onitoring Earth’s dynamic systems\, yet integrating multi-source\, multi
 -temporal data into globally consistent\, high-resolution representations 
 remains a challenge. Traditional remote sensing vision models\, which proc
 ess patches or images as inputs\, often struggle to capture fine-grained s
 patiotemporal-spectral relationships critical for downstream tasks like la
 nd classification\, climate modeling\, and change detection. We present a 
 self-supervised framework leveraging Barlow Twins to train an Earth Founda
 tion Model that outputs pixel-level representations from diverse satellite
  data sources. Unlike conventional ML approaches\, our model treats pixels
  as primary units of learning\, explicitly optimizing for temporal-spectra
 l coherence across billions of global 10m-resolution pixels. Preliminary r
 esults demonstrate that the resulting representation map encodes high-qual
 ity spatiotemporal patterns\, outperforming traditional ML methods in land
  classification. By bridging multi-modal satellite data into a harmonized 
 latent space\, our approach unlocks new opportunities for monitoring plane
 tary-scale processes with higher precision.\n\n*Bio*\n\nFrank Feng is a fi
 rst-year PhD student in the Department of Computer Science and Technology 
 at the University of Cambridge. His research interests lie at the intersec
 tion of machine learning and earth sciences\, with a particular focus on t
 he application of self-supervised learning in remote sensing.
LOCATION:GS15\, William Gates Building. Zoom link: https://cl-cam-ac-uk.zo
 om.us/j/4361570789?pwd=Nkl2T3ZLaTZwRm05bzRTOUUxY3Q4QT09&amp\;from=addon 
END:VEVENT
END:VCALENDAR