Skip to Main content Skip to Navigation
Conference papers

CHANNEL-BASED ATTENTION FOR LCC USING SENTINEL-2 TIME SERIES

Abstract : Deep Neural Networks (DNNs) are getting increasing attention to deal with Land Cover Classification (LCC) relying on Satellite Image Time Series (SITS). Though high performances can be achieved, the rationale of a prediction yielded by a DNN often remains unclear. An architecture expressing predictions with respect to input channels is thus proposed in this paper. It relies on convolutional layers and an attention mechanism weighting the importance of each channel in the final classification decision. The correlation between channels is taken into account to set up shared kernels and lower model complexity. Experiments based on a Sentinel-2 SITS show promising results.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03182049
Contributor : Hermann Courteille <>
Submitted on : Tuesday, March 30, 2021 - 10:17:39 AM
Last modification on : Thursday, April 8, 2021 - 1:45:39 PM

Files

SDeepIGARSS_final_to_submit25M...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03182049, version 1
  • ARXIV : 2103.16836

Citation

Hermann Courteille, A Benoit, N Méger, A Atto, D. Ienco. CHANNEL-BASED ATTENTION FOR LCC USING SENTINEL-2 TIME SERIES. International Geoscience and Remote Sensing Symposium (IGARSS), Jul 2021, Brussels, Belgium. ⟨hal-03182049⟩

Share

Metrics

Record views

48

Files downloads

9