Plotly Volume组件无法渲染随机分布3D顶点问题咨询及数据适配疑问

First off, let's get to the core issue: Plotly's Volume trace doesn't work with unstructured scatter points—even if you sort them. This component is designed for regular 3D Cartesian grids (structured data where each dimension has a fixed number of evenly spaced points, forming orthogonal grid lines). Randomly generated points lack this grid structure, so Plotly can't interpret them as a continuous volume to render isosurfaces.

What's Wrong with Your Code? #

When you use np.random.uniform to generate X/Y/Z, you're creating unstructured scatter data. Sorting these points doesn't fix the problem because they still don't form a regular grid. The official example relies on np.mgrid precisely because it generates this structured grid (each axis has a fixed step size and count, so every combination of X/Y/Z points forms a predictable cube-like structure).

How to Fix It #

Option 1: Use Regular Grid Data (Works Directly) #

Stick with np.mgrid to generate your coordinates, just like the official example. Here's how to adapt your logic:

import numpy as np
import plotly.graph_objects as go

# Generate a regular 3D grid with fixed steps
X, Y, Z = np.mgrid[0:1:100j, 0:1:100j, 0:1:100j]
# Calculate values across the grid
V = np.sin(X) * np.sin(Y) + Z

fig = go.Figure(data=go.Volume(
    x=X.flatten(), y=Y.flatten(), z=Z.flatten(),
    value=V.flatten(),
    isomin=np.min(V), isomax=np.max(V),
    opacity=0.1,
    surface_count=20,
    colorscale='Spectral', reversescale=True
))
fig.show()

Option 2: Interpolate Scatter Data to a Regular Grid #

If your third-party data is unstructured scatter points (shape (N,3) with corresponding values), you need to interpolate them onto a regular grid first. Use scipy.interpolate.griddata for this:

import numpy as np
import plotly.graph_objects as go
from scipy.interpolate import griddata

# Your unstructured scatter data (example)
np.random.seed(42)
coords = np.random.uniform(0, 1, (30000, 3))
X_scatter, Y_scatter, Z_scatter = coords[:,0], coords[:,1], coords[:,2]
V_scatter = np.sin(X_scatter) * np.sin(Y_scatter) + Z_scatter

# Create a regular grid to interpolate onto
xi, yi, zi = np.mgrid[0:1:50j, 0:1:50j, 0:1:50j]

# Interpolate scatter values onto the grid
V_interp = griddata((X_scatter, Y_scatter, Z_scatter), V_scatter, (xi, yi, zi), method='linear')

# Plot the interpolated volume
fig = go.Figure(data=go.Volume(
    x=xi.flatten(), y=yi.flatten(), z=zi.flatten(),
    value=V_interp.flatten(),
    isomin=np.nanmin(V_interp), isomax=np.nanmax(V_interp),
    opacity=0.1,
    surface_count=20,
    colorscale='Spectral', reversescale=True
))
fig.show()

Note: You might get NaN values in areas with no scatter points—you can handle these by filtering or switching to the nearest interpolation method.

Where to Host Your NumPy Dataset? #

For sharing NumPy .npy datasets:

• GitHub Gist: Quick, free, and great for small to medium datasets. You can upload .npy files directly or include code to regenerate the data.
• Kaggle Datasets: Ideal for larger datasets or data science-related projects; easy to share and discover.
• Zenodo: Perfect for academic datasets, as it provides DOIs for formal citation.
• Stack Overflow Question: For small datasets, convert the array to a CSV (using np.savetxt('data.csv', coords, delimiter=',')) and paste the content in a code block.

内容的提问来源于stack exchange，提问作者Simon T.