5 papers
UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation
Yohann Perron, Guillaume Astruc, Nicolas Gonthier +2
Vision Transformers (ViT) dominate computer vision. However, their reliance on rigid patch projectors hinders transfer to Earth Observation (EO), where input modalities, scales, an…
PoM: A Linear-Time Replacement for Attention with the Polynomial Mixer
David Picard, Nicolas Dufour, Lucas Degeorge +14
This paper introduces the Polynomial Mixer (PoM), a novel token mixing mechanism with linear complexity that serves as a drop-in replacement for self-attention. PoM aggregates inpu…
Adapting Vision Transformers to Ultra-High Resolution Semantic Segmentation with Relay Tokens
Yohann Perron, Vladyslav Sydorov, Christophe Pottier +1
Current approaches for segmenting ultra high resolution images either slide a window, thereby discarding global context, or downsample and lose fine detail. We propose a simple yet…
Archaeoscape: Bringing Aerial Laser Scanning Archaeology to the Deep Learning Era
Yohann Perron, Vladyslav Sydorov, Adam P. Wijker +3
Airborne Laser Scanning (ALS) technology has transformed modern archaeology by unveiling hidden landscapes beneath dense vegetation. However, the lack of expert-annotated, open-acc…
Open-Canopy: Towards Very High Resolution Forest Monitoring
Fajwel Fogel, Yohann Perron, Nikola Besic +8
Estimating canopy height and its changes at meter resolution from satellite imagery is a significant challenge in computer vision with critical environmental applications. However,…