Module: hyper_spectral_process.plot. Functions draw into a given axes or
return a figure; nothing is shown or saved implicitly (except the save_*
helpers).
| function | what it draws |
|---|---|
plot_map(ax, xy, values, cmap, vmin, vmax, percentiles, label, layout) |
one value per spectrum on the stage raster (layout="grid": cells of the raster spanned by the coordinates, missing nodes and NaN in grey; "scatter": square markers at the coordinates, for scans that are not rectangular). percentiles=(2, 98) clips the colour range to robust limits |
band_intensity(ssm, xmin, xmax, aggregate) |
sum, mean or max of every spectrum over an axis range ((N,) array) |
plot_band_intensity(ssm, xmin, xmax, aggregate, cmap="CMRmap", ax=None) |
the integrated-intensity map of a range (the original heatmap tool) |
plot_variant_map(ax, selection, xy) |
discrete map of the variant chosen per pixel |
Any per-pixel array can be mapped, for example a ratio of fitted areas:
from hyper_spectral_process.plot import plot_map
area = fit.result.peak_areas()
plot_map(ax, fit.result.xy, area[:, 1] / area[:, 0], cmap="RdBu_r", percentiles=(2, 98), label="area ratio")| function | what it draws |
|---|---|
plot_fit(ax, fsm, idx, residual=True, names=None) |
data, fit and every peak (filled above the baseline) of one row; residual offset below; inactive composite slots are left out |
plot_initial_guess(guess) |
the averaged spectrum with the model before and after the initial-guess fit, the refined peaks and the loss history |
plot_loss_history(ax, histories, stage_ends) |
mean loss per iteration over batches with a +/- 1 SD band and the stage boundaries |
save_summary_plots(fsm, out_dir) writes:
| file | content |
|---|---|
parameter_maps.png |
per peak: position, width, height and area maps |
quality_maps.png |
NRMSE, R^2 and baseline maps |
parameter_histograms.png |
per peak: histograms with the median |
average_fit.png |
mean data and mean fit over the fitted pixels with the mean residual |
sample_fits.png, worst_fits.png |
16 random fits; the 12 fits with the largest NRMSE |
ratio_maps.png |
area ratios of peak pairs |
position_height_scatter.png |
height against position, coloured by width |
save_variant_plots(selection, fits, xy, out_dir) writes the winner map with
the BIC-margin map and the BIC and wRMSE histograms, the mean spectrum and
fit over the pixels each variant wins, and (with a drift penalty) the
drift maps. Examples of both are in examples.