Python: Plotly 3D Surface Plot

Question

I am having a tough time drawing the Plotly 3d surface plot. I have a big data frame of 4000 rows and three columns. I did ask questions here and got some answers. When I try them, it takes hours for the code to run yet I see no plot. I want to confirm what I am doing is right. Because I am new to the surface plots.

My code:

import plotly.graph_objects as go
import plotly.graph_objs
import plotly
df = 
index     x           y           z
0        10.2        40.5        70.5            
1        30.5        30.2       570.5
.
.
4000     100.5       201.5      470.5

df['z']= [df['z'].tolist for x in df.index]
df = 
index     x           y           z
0        10.2        40.5       [70.5,570.5,..,470.5]            
1        30.5        30.2       [70.5,570.5,..,470.5]
.
.
4000     100.5       201.5      [70.5,570.5,..,470.5]

    zdata = [df['z'].tolist()]*len(df)
    plotly.offline.plot({"data":[go.Surface(x=df['x'].values,
                                            y=df['y'].values,
                                            z = df['z'].values)],
    "layout":plotly.graph_objs.Layout(title='Some data', autosize=False,
                      width=600, height=600,
                    scene = dict(xaxis_title='x',
                    yaxis_title='y',
                    zaxis_title='z'),
                      margin=dict(l=10, r=10, b=10, t=10))})

I would be grateful to have somebody clarify me that what I am doing to generate a surface plot is correct?

s3dev · Accepted Answer · 2020-08-11 20:07:57Z

4

Here is a simple / stripped down example of a 3D surface plot to hopefully get you going.

The key message here is: Don't over complicate it. This same logic should be fine on a DataFrame with 4000+ rows. (Granted, it'll plot ~16M data points, so it'll take a bit of time).

The key point to remember is that z must be a 2d array of the shape [x.shape[0], y.shape[0]]. Essentially meaning, if x and y are of length 10, then z must be of the shape: [10, 10].

As I don't have your complete dataset, I've synthesised the data - hope that's OK for illustration purposes. Additionally, I've stuck with numpy for simplicity, keeping in mind that a numpy array is essentially a DataFrame column.

Simple example:

import numpy as np
from plotly.offline import plot

n = 10
x = np.arange(n)
y = x
z = np.tile(x**2, [n, 1])

data = [{'x': x,
         'y': y,
         'z': z,
         'type': 'surface'}]

plot({'data': data}, filename='/path/to/graph.html')

Output:

Something a little more fun:

n = 360
x = np.arange(n)
y = x
v = np.tile([np.sin(i*(np.pi/180)) for i in range(n)], [n, 1]).T
z = (v.T[0]*v)

data = [{'x': x,
         'y': y,
         'z': z,
         'type': 'surface'}]

plot({'data': data}, filename='/path/to/graph.html')

You'll note the plotting logic is identical.

Output:

edited Aug 11, 2020 at 20:07

answered Aug 11, 2020 at 20:02

s3dev

9,8713 gold badges34 silver badges49 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

Mainland Over a year ago

How do I conver my data in z column to be of 2-D array of required shape? Is this same as what I have done already? or something different? Can you explain from this perspective?

s3dev Over a year ago

If all arrays are of the same length: A simple way would be: z = np.tile(x, [x.shape[0], 1]). This will 'tile' (or replicate) the x array into a shape of [10, 10] if x is of the length 10. This is the method I used in the example. However, typically with a 3D surface plot, you'll have x*y (in terms of array length) of z data points. If z is of the shape x*y: Use np.reshape([x.shape[0], y.shape[0]]) to reshape the z array, using the length of x and y.

Mainland Over a year ago

It is still confusing to me. The way I modified column z in my question, is it correct or wrong? I mean, I modified it to be x and y size? Right

s3dev Over a year ago

Note that using to_numpy() rather than tolist() should be faster for you. (Just ran on 1000 records and was ~6x faster).

s3dev Over a year ago

As simple as: df['z'] = [df['z'].to_numpy() for _ in df.index]

|

16807 · Accepted Answer · 2023-03-23 16:47:47Z

0

When you use go.Surface, z should be a 2-dimensional matrix, and x and y should be unique values of x_axis and y_axis. This code prepares DataFrame values for using go.Surface.

x_data = df['x'].values
y_data = df['y'].values
z_data = df['z'].values

x = np.unique(x_data)
y = np.unique(y_data)

# Set default value of np.Nan for z matrix
z = np.empty((x.shape[0], y.shape[0]))
z[:] = np.NaN
for i in range(len(z_data)):
    z[np.where(x, np.isclose(x, x_data[i])), 
      np.where(y, np.isclose(y, y_data[i]))] = z_data[i]
z=z.transpose()

fig = go.Figure(data=[go.Surface(z=z, x=x, y=y)])
fig.show()

edited Mar 23, 2023 at 16:47

16807

1,5442 gold badges20 silver badges33 bronze badges

answered Jul 19, 2022 at 16:59

Saeed Karimi

12 bronze badges

3 Comments

NoamG Over a year ago

in z_data[I], it seems like a typo - should be lowercase 'i'

NoamG Over a year ago

Anyway, it seems to to work anyway, it fails in the for-loop, where x_data[i] - x_min is not an integer.

16807 Over a year ago

I've updated the example to address the issue that @NoamG was having, since I was having that problem as well. The revised solution uses np.where(x, np.isclose(x, x_data[i])) to find indices within the arrays returned by np.unique. This should work for both floats and ints

Collectives™ on Stack Overflow

Python: Plotly 3D Surface Plot

2 Answers 2

Simple example:

Output:

Something a little more fun:

Output:

7 Comments

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Simple example:

Output:

Something a little more fun:

Output:

7 Comments

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related