Skip to content

Latest commit

 

History

History
98 lines (87 loc) · 4.46 KB

plots.md

File metadata and controls

98 lines (87 loc) · 4.46 KB
Raw

icoCNN.plots

Plotting functions for icosahedral signals.

icosahedral_charts

Plot the 5 charts of an icosahedral signal as a 5*2^r x 2^(r+1) image. The image is hard to interpret visually, but it's the projection used to perform the convolutions.

Parameters

  • x : torch tensor or numpy ndarray with shape [5, 2^r, 2^(r+1)] or [5, 2^r+2, 2^(r+1)+2] (padded signals), Icosahedral signal to plot
  • ax : matplotlib axes, optional, Axes to plot the signal in. It can be useful to make subplots. If it is not provided, a new matplotlib figure is created.
  • cmap : matplotlib cmap (optional), Colormap to use in the plot. Default is 'inferno'.
  • colorbar : bool (optional), Include a colorbar into the plot, default is False.

SphProjector

Class to plot icosahedral signals as spherical equiangular projections. This 2D figure easier to interpret, but preparing the projection is a bit slow. After having created the object, you can use its plot_projection method to plot signals.

Parameters

  • r : int, Resolution of the icosahedral signals that will be projected
  • res_theta : int (optional), Number of elevation points of the projections. Default is 512
  • res_phi : int (optional,) Number of azimuth points of the projections. Default is 1024

get_projection

Compute the projection of an icosahedral signal

Parameters
  • x : torch tensor or numpy ndarray with shape [5, 2^r, 2^(r+1)], Icosahedral signal to plot
Returns

Numpy ndarray with shape res_theta x res_phi+1, The last column of the matrix will be equal to the first one since both represent phi = -pi = +pi.

plot_projection

Plot the projection of an icosahedral signal

Parameters
  • x : torch tensor or numpy ndarray with shape [5, 2^r, 2^(r+1)], Icosahedral signal to project
  • ax : matplotlib axes (optional), Axes to plot the signal in. It can be useful to make subplots. If it is not provided, a new matplotlib figure is created
  • cmap : matplotlib cmap (optional), Colormap to use in the plot. Default is 'inferno'
  • colorbar : bool (optional), Include a colorbar into the plot, default is False

icosahedral_scatter

Plot an icosahedral signal as a 3D scatter plot where each hexagonal pixel is a 3D point with the proper color. The visual result isn't very good, but it's easy to do, so it can be useful for some quick tests. icoCNN.plots.draw_icosahedron provides better 3D plots using VPython.

Parameters

  • x : torch tensor or numpy ndarray with shape [5, 2^r, 2^(r+1)], Icosahedral signal to plot
  • grid : 4D ndarray with shape [5, 2^r, 2^(r+1), 3] (optional), 3D Cartesian coordinates of every point of the icosahedral grid where x is defined. If it is not provided, it is computed inferring its resolution from the shape of x
  • ax : matplotlib axes (optional), Axes to plot the signal in. It should have projection='3d'. If it is not provided, a new matplotlib figure is created
  • cmap : matplotlib cmap (optional), Colormap to use in the plot. Default is 'inferno'

draw_icosahedron

Draw an icosahedral signal in 3D using VPython. The vertices of each hexagonal pixel is computed according to their neighbors, so they aren't accurate at the edges. This effect is negligible when using icosahedral signals with resolution r=3 but is visible for lower resolutions.

Parameters

  • x : torch tensor or numpy ndarray with shape [5, 2^r, 2^(r+1)], Icosahedral signal to draw
  • canvas : VPython canvas (optional), Canvas to draw the signal in. If it is not provided, the active one will be used (or a new one will be created)