2

I have a large dataset of the form [(X1, Y1, Z1, VALUE1), (X2, Y2, Z2, VALUE2)...]. The geometry of the points is the surface of a cylinder, while there are many discrete points they come nowhere near being a full mesh.

I would like to create a basic plot, where each of the points is given an intensity of a color (like a heatmap) based on how high its value is, and then the colors are smoothed to some degree to create a cohesive surface rather than discrete points

I am currently using matplotlib, however, I would also use other libraries if necessary.

I have looked into both surface plots and Tri-Surface plots but neither seem to do what I want (although the documentation for plot_trisurf() is a little confusing so maybe it is still a possibility). I have also looked at this post: 3D discrete heatmap in matplotlib.

And while the set up is mostly the same, I would like to have a more cohesive surface plot rather than a 3d Tetris set up. The original answer seems pretty close to my desired solution, however, I would like the colors to be based on VALUE rather than Z and if possible for there to be color smoothing between the sections.

Patrick Jordan
  • 33
  • 1
  • 4
  • Bokeh might have some good examples that you could reference for charting. [Link to samples](https://bokeh.pydata.org/en/latest/docs/gallery.html) – Kyle Jul 09 '18 at 21:56
  • Try Mayavi. http://docs.enthought.com/mayavi/mayavi/auto/examples.html https://www.lfd.uci.edu/~gohlke/pythonlibs/#mayavi – Joe Jul 10 '18 at 06:11

1 Answers1

3

Depending on how dense your point cloud is you may be able to get what you want with this (adjust the size parameter, s, to fill out the plot best for your data):

from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure()
ax = fig.gca(projection='3d')
ax.scatter(X, Y, Z, c=Value, lw=0, s=20)
plt.show()

enter image description here

screenpaver
  • 1,120
  • 8
  • 14
  • 1
    This is great, thanks so much. While it definitely looks mesh-like, is there any way to force the points to connect and have some sort of color smoothing between them? – Patrick Jordan Jul 10 '18 at 17:26
  • Not that I know of, with the above approach. When I work with cylinders I sometimes create radius and angle variables from the X,Y data, create a regular grid (Z vs angle) with numpy.linspace, and then use scipy griddata to interpolate my data onto the grid. Then I use contourf to plot continuous contours on a rolled out cylinder (flat rectangle). – screenpaver Jul 10 '18 at 19:05
  • @screenpaver could you share your code, i need also something similar but till now could not get any result as your. thanks –  Aug 30 '19 at 12:23