Laboratory for the use of algorithms in HyperSpectral imagery.

• Anaconda
• Git

This repository contains the following submodules:

• IPDL

In order to download the submodules in the cloning process, use the following instruction:

git clone --recurse-submodules [email protected]:SolidusAbi/HySpecLab.git

1. PyTorch

   • version 1.10 or above

2. Torchvision

   • version 0.11 or above

3. Statsmodels

   • version 0.12.2

4. Scikit-Learn

   • version 0.23
   • Recommended to install scikit-learn-intelex package

5. Scikit-Optimize

   • version 0.8.1

6. Imbalanced learn

   • version 0.7

7. Spectral

   • version 0.22

8. PySpTools

   • For Endmember Extraction Algorithms.
   • version 0.15

9. Others:

   • Matplotlib, tensorboard, tqdm...

If you have included a new library in your environment, please update the environment using the following command:

conda env export --from-history | grep -v "^prefix: " > environments/Linux.yml

TODO

• Conda Cheat Sheet

• SVDD

• Setup library.
• Fix HyperSpectralUnderSampler.
• Use VCA as endmember extraction algorithm.
• Extensive testing of the VCA with the original code.
• UnDIP documentation.
• NFINDR implementation based on pysptools. Básicamente para cargarme la dependencia a PySpTools.
• Experimenting with UnDIP with different data transformation.
• Endmember extraction by SVDD
  • Proposed on 10.1109/IGARSS.2007.4423675
• SparseWeightedFeatureSelector
  • Move the KL implementation to Sparse library

• Feature selection experiment
  • Plastic dataset, algorithms and sparse representation test
• Subspace SVDD
  • Quizás es la solución al mal performance que da el SVDD como eea pero... puedes ser demasiado complejo.
  • Paper