Plotting and Analysis¶
The ava.plotting subpackage contains useful functions for plotting and
analysis. These are very easy to use if you’ve already made a DataContainer
object (see previous section):
from ava.plotting.grid_plot import indexed_grid_plot_DC
dc = ... # define DataContainer
indices = [[0,1,2], [3,4,5]]
indexed_grid_plot_DC(dc, indices)
This plots a 2-by-3 grid of spectrograms with indices determined by
indices and saves the image to
os.path.join(dc.plots_dir, 'grid.pdf').
Try help(ava.plotting.grid_plot) or read the docs for more options.
Another useful plot:
from ava.plotting.latent_projection import latent_projection_plot_DC
dc = ... # define DataContainer
latent_projection_plot_DC(dc, embedding_type='latent_mean_umap')
This plots a UMAP projection of the latent means and saves the result to
os.path.join(dc.plots_dir, 'latent.pdf').
See the ava.plotting documentation for more plotting and analysis tools.
Shotgun VAE¶
In order to use some of these plotting functions with the shotgun VAE, we need
to first save some spectrograms. The shotgun VAE dataset has a method to do
this: write_hdf5_files
spec_dir = 'where/to/save/specs'
loaders['test'].dataset.write_hdf5_files(spec_dir, num_files=1000)
Warped Shotgun VAE¶
TO DO