如何调整分组箱线图的尺寸以清晰显示标签？

The Problem #

I'm using the following code to create a grouped boxplot with Seaborn and Matplotlib:

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import os
import matplotlib.font_manager as font_manager
import sys
import numpy as np
import matplotlib as mpl

path = os.getcwd() + "/results/"
sns.set_theme(style="whitegrid")
df = pd.read_csv("C:/tmp/all2.txt")
ax = sns.boxplot(x="cluster", y="val", hue="type", palette=["k", "w"], data=df,showfliers = False)
#sns.despine(ax=ax, trim=True, offset={'left':1,'right':1,'top':1,'bottom':1})
ax.set(ylabel='Number', xlabel='Clustered profiles')
font = font_manager.FontProperties(family='sans-serif', weight='bold', style='normal')
plt.legend(loc='best', frameon=False, prop=font)
plt.legend(loc='best', frameon=False, prop=font)
plt.xticks(weight='bold', fontname='sans-serif')
plt.yticks(weight='bold', fontname='sans-serif')
plt.xlabel("Clustered profiles", weight='bold', fontname='sans-serif', size=14)
plt.tight_layout()
plt.savefig(path + "/myoutput.pdf", dpi=250, transparent=False, bbox_inches='tight', format="pdf")

Some labels in the generated plot are unreadable. I tried increasing the dpi when saving, but it didn't help. What settings am I missing?

Solutions to Fix Blurry Labels #

Let's walk through the key adjustments to make your labels sharp and readable:

1. Set Global Font & Rendering Defaults #

Blurry text often comes from small default font sizes or disabled anti-aliasing. Add these lines at the start of your script to configure global Matplotlib settings:

# Configure global matplotlib settings for better text rendering
mpl.rcParams['font.size'] = 12  # Base font size for all elements
mpl.rcParams['axes.labelsize'] = 14  # Default axis label size
mpl.rcParams['xtick.labelsize'] = 12  # X-axis tick label size
mpl.rcParams['ytick.labelsize'] = 12  # Y-axis tick label size
mpl.rcParams['text.antialiased'] = True  # Enable anti-aliasing for smoother text

2. Remove Duplicate Legend Calls #

You’re calling plt.legend() twice, which is redundant and can cause unexpected behavior. Replace both calls with a single, properly configured legend:

# Single legend call with explicit font settings
plt.legend(
    loc='best', 
    frameon=False, 
    prop={'family':'sans-serif', 'weight':'bold', 'size':12}
)

3. Standardize Font Sizes for All Text Elements #

Your current code sets the x-label size to 14 but leaves the y-label and tick labels without explicit size settings. Update these to ensure consistency:

# Set axis labels with matching size and weight
plt.xlabel("Clustered profiles", weight='bold', fontname='sans-serif', size=14)
plt.ylabel("Number", weight='bold', fontname='sans-serif', size=14)

# Set tick labels with explicit font size
plt.xticks(weight='bold', fontname='sans-serif', fontsize=12)
plt.yticks(weight='bold', fontname='sans-serif', fontsize=12)

4. Verify Font Availability #

If the sans-serif font family isn’t mapped to a clear, readable font on your system, labels might render poorly. Try specifying a specific font like Arial (if available on your machine):

font = font_manager.FontProperties(family='Arial', weight='bold', style='normal', size=12)

5. Understand Vector vs. Raster Output #

Since you’re saving as a PDF (a vector format), increasing DPI won’t improve text sharpness—vector graphics scale infinitely without losing quality. The core issue is text size, not resolution. Focus on adjusting font sizes rather than DPI for PDF outputs.

Revised Full Code #

Here’s the complete updated code incorporating all fixes:

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import os
import matplotlib.font_manager as font_manager
import numpy as np
import matplotlib as mpl

# Configure global matplotlib settings for sharp, readable text
mpl.rcParams['font.size'] = 12
mpl.rcParams['axes.labelsize'] = 14
mpl.rcParams['xtick.labelsize'] = 12
mpl.rcParams['ytick.labelsize'] = 12
mpl.rcParams['text.antialiased'] = True

path = os.getcwd() + "/results/"
sns.set_theme(style="whitegrid")
df = pd.read_csv("C:/tmp/all2.txt")

ax = sns.boxplot(
    x="cluster", 
    y="val", 
    hue="type", 
    palette=["k", "w"], 
    data=df,
    showfliers=False
)

# Set axis labels with consistent styling
plt.xlabel("Clustered profiles", weight='bold', fontname='sans-serif', size=14)
plt.ylabel("Number", weight='bold', fontname='sans-serif', size=14)

# Configure legend with proper font settings
plt.legend(
    loc='best', 
    frameon=False, 
    prop={'family':'sans-serif', 'weight':'bold', 'size':12}
)

# Style tick labels
plt.xticks(weight='bold', fontname='sans-serif', fontsize=12)
plt.yticks(weight='bold', fontname='sans-serif', fontsize=12)

plt.tight_layout()
# Save as PDF (vector format) – high DPI isn't necessary here
plt.savefig(
    path + "/myoutput.pdf", 
    transparent=False, 
    bbox_inches='tight', 
    format="pdf"
)

内容的提问来源于stack exchange，提问作者Cenk Ten