spectraclass

Jupyterlab workbench supporting visual exploration and classification of high dimensional sensor data.

Conda CPU Environment Setup

> conda create --prefix /explore/nobackup/projects/ilab/conda/envs/spectraclass -c conda-forge python=3.9 mamba
> conda activate spectraclass
> mamba install -c conda-forge ipympl jupytext pyepsg ipysheet tensorflow h5py pythreejs nb_conda_kernels nodejs jupyterlab jupyterlab_server ipywidgets numpy xarray matplotlib rasterio scipy scikit-learn dask netcdf4 scikit-image numba gdal owslib rioxarray cartopy shapely bottleneck geopandas 

> conda create --name spectraclass.hv -c pyviz -c conda-forge mamba holoviews geopandas geoviews hvplot
> conda create --prefix /explore/nobackup/projects/ilab/conda/envs/spectraclass.hv -c pyviz -c conda-forge mamba holoviews geopandas geoviews hvplot

> conda activate spectraclass.hv
> mamba install -c pyviz -c conda-forge ipympl jupytext pyepsg ipysheet tensorflow h5py pythreejs nb_conda_kernels nodejs jupyterlab jupyterlab_server  rasterio dask netcdf4 scikit-image numba owslib rioxarray bottleneck  

The x-ray application requires the following additional packages:

> mamba install -c conda-forge jupyter_bokeh

Installation

$ git clone https://github.com/nasa-nccs-cds/spectraclass.git
$ cd spectraclass
$ pip install .

Image Index Creation

For example, with DESIS data:

gdaltindex -t_srs EPSG:32618 image_index_srs.shp *-SPECTRAL_IMAGE.tif

When actively developing your extension, build Jupyter Lab with the command:

$ jupyter lab --watch

This takes a minute or so to get started, but then automatically rebuilds JupyterLab when your javascript changes.

Note on first jupyter lab --watch, you may need to touch a file to get Jupyter Lab to open.