Develop Reference

Add image annotations to boxplot

I would like to add image annotations to a boxplot, akin to what they did with the bar chart in this post:
How can I add images to bars in axes (matplotlib)
My dataframe looks like this:
import pandas as pd
import numpy as np
names = ['PersonA', 'PersonB', 'PersonC', 'PersonD','PersonE','PersonF']
regions = ['NorthEast','NorthWest','SouthEast','SouthWest']
dates = pd.date_range(start = '2021-05-28', end = '2021-08-23', freq = 'D')
df = pd.DataFrame({'runtime': np.repeat(dates, len(names))})
df['name'] = len(dates)*names
df['A'] = 40 + 20*np.random.random(len(df))
df['B'] = .1 * np.random.random(len(df))
df['C'] = 1 +.5 * np.random.random(len(df))
df['region'] = np.resize(regions,len(df))
I tried to use the AnnotationBbox method which worked great for my time-series, but I'm not entirely sure if it can be applied here.
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.offsetbox import OffsetImage, AnnotationBbox
from matplotlib.cbook import get_sample_data
fig, ax = plt.subplots(
df.boxplot(column='A', by=['name'],ax=ax,showmeans=True, fontsize=8, grid=False)
for name in names:
rslt_df = df[df['name']==name]
val = rslt_df['A'].values[0]
xy = (0, val)
fn = get_sample_data(f"{name}.png", asfileobj=False)
arr_img = plt.imread(fn, format='png')
imagebox = OffsetImage(arr_img, zoom=0.125)
imagebox.image.axes = ax
ab = AnnotationBbox(imagebox, xy,xybox=(15.,0),xycoords='data',boxcoords="offset points",pad=0,frameon=False)
ax.add_artist(ab)

The code in the OP if very similar to Add image annotations to bar plots axis tick labels, but needs to be modified because boxplots are slightly different the barplots.
The main issue was xy didn't have the correct values.
The xy and xybox parameters can be adjusted to place the images anywhere.
By default, boxplot positions the ticks at range(1, n+1), as explained in this answer
Reset the tick positions with a 0 index: positions=range(len(names))
df was created with names = ['PersonA', 'PersonB', 'PersonC'] since only 3 images were provided.
ax = df.boxplot(column='A', by=['name'], showmeans=True, fontsize=8, grid=False, positions=range(len(names)))
ax.set(xlabel=None, title=None)
# move the xtick labels
ax.set_xticks(range(len(names)))
ax.set_xticklabels(countries)
ax.tick_params(axis='x', which='major', pad=30)
# use the ytick values to locate the image
y = ax.get_yticks()[1]
for i, (name, data) in enumerate(df.groupby('name')):
xy = (i, y)
fn = f"data/so_data/2021-08-28/{name}.png" # path to file
arr_img = plt.imread(fn, format='png')
imagebox = OffsetImage(arr_img, zoom=0.125)
imagebox.image.axes = ax
ab = AnnotationBbox(imagebox, xy, xybox=(0, -30), xycoords='data', boxcoords="offset points", pad=0, frameon=False)
ax.add_artist(ab)

I noticed that the y-ticks don't always position themselves in a friendly manner, so I set a static Y-value (the x-axis). Creating a transform xycoords, allows placement directly below the x-axis, no matter the y-tick scale.
# BOX GRAPH PLOT
fig, ax = plt.subplots(facecolor='darkslategrey')
plt.style.use('dark_background')
ax = df.boxplot(column=str(c), by=['name'],ax=ax,showmeans=True, fontsize=8,grid=False,positions=range(len(top)))
ax.set(xlabel=None, title=None)
# move the xtick labels
ax.set_xticks(range(len(top)))
ax.tick_params(axis='x', which='major', pad=20)
# use the ytick values to locate the image
y = ax.get_xticks()[0]
for i, (name, data) in enumerate(df.groupby('name')):
xy = (i, y)
fn = f"{imgsrc}/{name}.png" # path to file
arr_img = plt.imread(fn, format='png')
imagebox = OffsetImage(arr_img, zoom=0.125)
imagebox.image.axes = ax
trans = ax.get_xaxis_transform()
ab = AnnotationBbox(imagebox, xy, xybox=(0, -15), xycoords=trans,boxcoords="offset points", pad=0, frameon=False)
ax.add_artist(ab)
plt.show()

Add Legend for scatter plot

Below is the code for scatter plot.
for_tsne = np.hstack((X_embedding, y.values.reshape(-1,1)))
for_tsne_df = pd.DataFrame(data=for_tsne, columns=
['Dimension_x','Dimension_y','Labels'])
colors = {0:'red', 1:'blue', 2:'yellow'}
#colors = ['red','blue']
plt.scatter(for_tsne_df['Dimension_x'],
for_tsne_df['Dimension_y'],c=for_tsne_df['Labels'].apply(lambda x:
colors[x]))
plt.title("TSNE with BOW encoding of project_title")
plt.xlabel("Dimension_x")
plt.ylabel("Dimension_y")
plt.legend()
plt.show()`
How can I add legend? Above code is displaying only one label as Dimension_y.

One option is to assign a label to plt.scatter(). The legend will only appear if you plot the data with a label:
import matplotlib.pyplot as plt
import numpy as np
x = np.random.random(size=(100))
y = np.random.random(size=(100))
x1 = np.random.random(size=(100))
y1 = np.random.random(size=(100))
plt.scatter(x,y, label='sample 1')
plt.scatter(x1,y1, label='sample 2')
plt.title("TSNE with BOW encoding of project_title")
plt.xlabel("Dimension_x")
plt.ylabel("Dimension_y")
plt.legend()
plt.show()

Python Matplotlib Twinx() cursor values

I want to use the cursor (x,y values get displayed at the bottom left of figure) to measure the y and x distance between two points, however this only works for the data plotted on the second axis.
Is there a way to switch back an forth between the second axis and first y-axis?
Please note: I do not want a programmatic way of getting distance between points, just to use the cursor when I am viewing data in the figure plot.
Not sure if this helps but my code is literally the example for plotting two axes from the matplotlib page:
fig, ax1 = plt.subplots()
ax1.plot(sensor1, 'b-')
ax1.set_xlabel('(time)')
# Make the y-axis label and tick labels match the line color.
ax1.set_ylabel('Sensor 1', color='b')
for tl in ax1.get_yticklabels():
tl.set_color('b')
ax2 = ax1.twinx()
ax2.plot(sensor2, 'r.')
ax2.set_ylabel('Sensor 2', color='r')
for tl in ax2.get_yticklabels():
tl.set_color('r')
plt.show()

You can use the excellent answer here to get both coordinates displayed at the same time. In order to get distance between two points, you can then combine this idea with ginput to map from one to the other and add the result as a title,
import matplotlib.pyplot as plt
import numpy as np
#Provide other axis
def get_othercoords(x,y,current,other):
display_coord = current.transData.transform((x,y))
inv = other.transData.inverted()
ax_coord = inv.transform(display_coord)
return ax_coord
#Plot the data
fig, ax1 = plt.subplots()
t = np.linspace(0,2*np.pi,100)
ax1.plot(t, np.sin(t),'b-')
ax1.set_xlabel('(time)')
ax1.set_ylabel('Sensor 1', color='b')
for tl in ax1.get_yticklabels():
tl.set_color('b')
ax2 = ax1.twinx()
ax2.plot(t,3.*np.cos(t),'r-')
ax2.set_ylabel('Sensor 2', color='r')
for tl in ax2.get_yticklabels():
tl.set_color('r')
#Get user input
out = plt.ginput(2)
#2nd axis from input
x2b, x2t = out[0][0], out[1][0]
y2b, y2t = out[0][1], out[1][1]
#Draw line
ax2.plot([x2b, x2t],[y2b, y2t],'k-',lw=3)
#1st axis from transform
x1b, y1b = get_othercoords(x2b,y2b,ax2,ax1)
x1t, y1t = get_othercoords(x2t,y2t,ax2,ax1)
plt.title("Distance x1 = " + str(x1t-x1b) + " y1 = " + str(y1t-y1b) + "\n"
"Distance x2 = " + str(x2t-x2b) + " y2 = " + str(y2t-y2b))
plt.draw()
plt.show()
which gives something like,

Matplotlib - set an axis label where there are no ticks

Hello Python/Matplotlib gurus,
I would like to label the y-axis at a random point where a particular horizontal line is drawn.
My Y-axis should not have any values, and only show major ticks.
To illustrate my request clearly, I will use some screenshots.
What I have currently:
What I want:
As you can see, E1 and E2 are not exactly at the major tick mark. Actually, I know the y-axis values (although they should be hidden, since it's a model graph). I also know the values of E1 and E2.
I would appreciate some help.
Let my code snippet be as follows:
ax3.axis([0,800,0,2500) #You can see that the major YTick-marks will be at 500 intervals
ax3.plot(x,y) #plot my lines
E1 = 1447
E2 = 2456
all_ticks = ax3.yaxis.get_all_ticks() #method that does not exist. If it did, I would be able to bind labels E1 and E2 to the respective values.
Thank you for the help!
Edit:
For another graph, I use this code to have various colors for the labels. This works nicely. energy_range, labels_energy, colors_energy are numpy arrays as large as my y-axis, in my case, 2500.
#Modify the labels and colors of the Power y-axis
for i, y in enumerate(energy_range):
if (i == int(math.floor(E1))):
labels_energy[i] = '$E_1$'
colors_energy[i] = 'blue'
elif (i == int(math.floor(E2))):
labels_energy[i] = '$E_2$'
colors_energy[i] = 'green'
else:
labels_energy.append('')
#Modify the colour of the energy y-axis ticks
for color,tick in zip(colors_energy,ax3.yaxis.get_major_ticks()):
print color, tick
if color:
print color
tick.label1.set_color(color) #set the color property
ax3.get_yaxis().set_ticklabels(labels_energy)
Edit2:
Full sample with dummy values:
#!/bin/python
import matplotlib
# matplotlib.use('Agg') #Remote, block show()
import numpy as np
import pylab as pylab
from pylab import *
import math
import matplotlib.pyplot as plt
from matplotlib.ticker import MultipleLocator
import matplotlib.font_manager as fm
from matplotlib.font_manager import FontProperties
import matplotlib.dates as mdates
from datetime import datetime
import matplotlib.cm as cm
from matplotlib.ticker import MultipleLocator, FormatStrFormatter
from scipy import interpolate
def plot_sketch():
x = np.arange(0,800,1)
energy_range = range (0,2500,1) #Power graph y-axis range
labels_energy = [''] * len(energy_range)
colors_energy = [''] * len(energy_range)
f1=4
P1=3
P2=2
P3=4
f2=2
f3=6
#Set Axes ranges
ax3.axis([0,800,0,energy_range[-1]])
#Add Energy lines; E=integral(P) dt
y=[i * P1 for i in x]
ax3.plot(x,y, color='b')
y = [i * P2 for i in x[:0.3*800]]
ax3.plot(x[:0.3*800],y, color='g')
last_val = y[-1]
y = [(i * P3 -last_val) for i in x[(0.3*800):(0.6*800)]]
ax3.plot(x[(0.3*800):(0.6*800)],y, color='g')
E1 = x[-1] * P1
E2 = (0.3 * x[-1]) * P2 + x[-1] * (0.6-0.3) * P3
#Modify the labels and colors of the Power y-axis
for i, y in enumerate(energy_range):
if (i == int(math.floor(E1))):
labels_energy[i] = '$E_1$'
colors_energy[i] = 'blue'
elif (i == int(math.floor(E2))):
labels_energy[i] = '$E_2$'
colors_energy[i] = 'green'
else:
labels_energy.append('')
#Modify the colour of the power y-axis ticks
for color,tick in zip(colors_energy,ax3.yaxis.get_major_ticks()):
if color:
tick.label1.set_color(color) #set the color property
ax3.get_yaxis().set_ticklabels(labels_energy)
ax3.axhline(energy_range[int(math.floor(E1))], xmin=0, xmax=1, linewidth=0.25, color='b', linestyle='--')
ax3.axhline(energy_range[int(math.floor(E2))], xmin=0, xmax=0.6, linewidth=0.25, color='g', linestyle='--')
#Show grid
ax3.xaxis.grid(True)
#fig = Sketch graph
fig = plt.figure(num=None, figsize=(14, 7), dpi=80, facecolor='w', edgecolor='k')
fig.canvas.set_window_title('Sketch graph')
ax3 = fig.add_subplot(111) #Energy plot
ax3.set_xlabel('Time (ms)', fontsize=12)
ax3.set_ylabel('Energy (J)', fontsize=12)
pylab.xlim(xmin=0) # start at 0
plot_sketch()
plt.subplots_adjust(hspace=0)
plt.show()

I think you're looking for the correct transform (check this out). In your case, what I think you want is to simply use the text method with the correct transform kwarg. Try adding this to your plot_sketch function after your axhline calls:
ax3.text(0, energy_range[int(math.floor(E1))],
'E1', color='g',
ha='right',
va='center',
transform=ax3.get_yaxis_transform(),
)
ax3.text(0, energy_range[int(math.floor(E2))],
'E2', color='b',
ha='right',
va='center',
transform=ax3.get_yaxis_transform(),
)
The get_yaxis_transform method returns a 'blended' transform which makes the x values input to the text call be plotted in axes units, and the y data in 'data' units. You can adjust the value of the x-data, (0) to be -0.003 or something if you want a little padding (or you could use a ScaledTranslation transform, but that's generally unnecessary if this is a one-off fix).
You'll probably also want to use the 'labelpad' option for set_ylabel, e.g.:
ax3.set_ylabel('Energy (J)', fontsize=12, labelpad=20)

I think my answer to a different post might be of help to you:
Matplotlib: Add strings as custom x-ticks but also keep existing (numeric) tick labels? Alternatives to matplotlib.pyplot.annotate?
It also works for the y-axis.Here is the result:

Setting the labels of a chart

I copy this example from matplotlib site and now I want to change the font, color and size of the labels, but not the numbers size. And there is a possiblity to just see the numbers that are in the middle and at the end of each side?
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import numpy as np
fig = plt.figure()
ax = fig.gca(projection='3d')
X, Y = np.mgrid[0:6*np.pi:0.25, 0:4*np.pi:0.25]
Z = np.sqrt(np.abs(np.cos(X) + np.cos(Y)))
surf = ax.plot_surface(X + 1e5, Y + 1e5, Z, cmap='autumn', cstride=2, rstride=2)
ax.set_xlabel("X-Label")
ax.set_ylabel("Y-Label")
ax.set_zlabel("Z-Label")
ax.set_zlim(0, 2)
plt.show()
Thank you

You can change the font size, color and type by setting text properties when you create the label like this:
ax.set_xlabel("X-Label", size = 40, color = 'r', family = 'fantasy')
You can control which tick marks are displayed using ax.set_xticks.