A novel end-to-end dropout concrete autoencoder for band selection on hyperspectral images outperforms prior methods on four scenes by training directly on the target band subset.
The concrete distribution: A continuous relaxation of discrete random variables,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Dropout Concrete Autoencoder for Band Selection on HSI Scenes
A novel end-to-end dropout concrete autoencoder for band selection on hyperspectral images outperforms prior methods on four scenes by training directly on the target band subset.