SpACNN-LDVAE: Spatial Attention Convolutional Latent Dirichlet Variational Autoencoder for Hyperspectral Pixel Unmixing

  • 2023-11-17 18:45:00
  • Soham Chitnis, Kiran Mantripragada, Faisal Z. Qureshi
  • 0


The Hyperspectral Unxming problem is to find the pure spectral signal of theunderlying materials (endmembers) and their proportions (abundances). Theproposed method builds upon the recently proposed method, Latent DirichletVariational Autoencoder (LDVAE). It assumes that abundances can be encoded asDirichlet Distributions while mixed pixels and endmembers are represented byMultivariate Normal Distributions. However, LDVAE does not leverage spatialinformation present in an HSI; we propose an Isotropic CNN encoder with spatialattention to solve the hyperspectral unmixing problem. We evaluated our modelon Samson, Hydice Urban, Cuprite, and OnTech-HSI-Syn-21 datasets. Our modelalso leverages the transfer learning paradigm for Cuprite Dataset, where wetrain the model on synthetic data and evaluate it on real-world data. We areable to observe the improvement in the results for the endmember extraction andabundance estimation by incorporating the spatial information. Code can befound at


Quick Read (beta)

loading the full paper ...