如何提取Seaborn散点图图例形状并分配给颜色分组？

• 目标：制作结合颜色与形状的Seaborn散点图，切换变量映射逻辑：不再将颜色分配给形状，而是将形状条目分配给颜色条目
• 当前问题：已实现「颜色分配给形状」的功能，但图表视觉混乱（蓝色占比过大）
• 期望调整：设置hue='Shorthand'、style='Subscale'，实现形状与颜色图例条目的关联

#subscale_dict = {'Deccan Traps': 'CFB', 'Other fissure': 'Fissure', 'Experimental': 'Experimental', 'Stratovolcano': 'Stratovolcano',
#                'Columbia river basalts': 'CFB', 'Iceland': 'Fissure', 'Shield': 'Shield', 'Ferrar': 'CFB', 'Emeishan': 'CFB', 
#                'EquiMAP': 'CFB', 'Moon': 'Moon'}
subscale_dict = passframe.set_index("Shorthand")["Style"].to_dict()
passframe['Subscale'] = passframe['Shorthand'].map(subscale_dict)
passframe['Subscale'] = pd.Categorical(passframe['Subscale'])

print(subscale_dict)

passframe
cust = {'axes.labelsize': 24, 'xtick.labelsize':20, 'ytick.labelsize':20, 'legend.fontsize':20, 'figure.facecolor': 'white'}
sns.set(rc={'figure.figsize':(32,18)})
sns.set_theme(style = 'ticks', font = 'Calibri', rc = cust)
sns.set_context('poster')
colour = 'colorblind'


fig1 = sns.scatterplot(data = passframe, x = 'Slope1', y = 'Intercept1', style = 'Shorthand',
                   hue = 'Subscale',s = 400, palette = colour, legend = 'full')#, height = 12, aspect = 16/9)

handles, labels = fig1.get_legend_handles_labels()
index_item_title = labels.index('Shorthand')

color_dict = {label: handle.get_facecolor()
              for handle, label in zip(handles[1:index_item_title], labels[1:index_item_title])}

print(color_dict)
# loop through the items, assign color via the subscale of the item idem
for handle, label in zip(handles[index_item_title + 1:], labels[index_item_title + 1:]):
    handle.set_color(color_dict[subscale_dict[label]])
    
fig1.legend(handles[index_item_title + 1:], labels[index_item_title + 1:], title='Data',
          loc = 'best', markerscale = 2, labelspacing = 1.2, fontsize = 24, title_fontsize = 32)


fig1.set_xlabel('$Slope (cm^{-1})$')
fig1.set_ylabel('$Intercept (cm^{-4})$')
plt.xlim([min(passframe['Slope1']), 1500])
plt.ylim([min(passframe['Intercept1']-2), max(passframe['Intercept2'])+2])

当前实现效果 #

核心调整：交换hue与style的映射变量，重构图例逻辑实现形状到颜色条目的分配

import seaborn as sns
import pandas as pd
import matplotlib.pyplot as plt

# 构建Subscale映射字典
subscale_dict = passframe.set_index("Shorthand")["Style"].to_dict()
passframe['Subscale'] = passframe['Shorthand'].map(subscale_dict)
passframe['Subscale'] = pd.Categorical(passframe['Subscale'])

# 绘图样式配置
cust = {'axes.labelsize': 24, 'xtick.labelsize':20, 'ytick.labelsize':20, 'legend.fontsize':20, 'figure.facecolor': 'white'}
sns.set(rc={'figure.figsize':(32,18)})
sns.set_theme(style = 'ticks', font = 'Calibri', rc = cust)
sns.set_context('poster')
colour = 'colorblind'

# 交换hue和style变量：Shorthand对应颜色，Subscale对应形状
fig1 = sns.scatterplot(data = passframe, x = 'Slope1', y = 'Intercept1', 
                       hue = 'Shorthand', style = 'Subscale',
                       s = 400, palette = colour, legend = 'full')

# 提取图例手柄与标签，调整形状与颜色的关联
handles, labels = fig1.get_legend_handles_labels()
# 定位Subscale图例的起始索引
index_style_title = labels.index('Subscale')

# 构建形状映射字典：Subscale对应标记形状
style_dict = {label: handle.get_marker()
              for handle, label in zip(handles[1:index_style_title], labels[1:index_style_title])}

# 遍历颜色条目（Shorthand），分配对应Subscale的形状
for handle, label in zip(handles[index_style_title + 1:], labels[index_style_title + 1:]):
    handle.set_marker(style_dict[subscale_dict[label]])

# 更新图例，仅保留Shorthand相关条目
fig1.legend(handles[index_style_title + 1:], labels[index_style_title + 1:], title='Data',
          loc = 'best', markerscale = 2, labelspacing = 1.2, fontsize = 24, title_fontsize = 32)

# 设置坐标轴标签和范围
fig1.set_xlabel('$Slope (cm^{-1})$')
fig1.set_ylabel('$Intercept (cm^{-4})$')
plt.xlim([min(passframe['Slope1']), 1500])
plt.ylim([min(passframe['Intercept1']-2), max(passframe['Intercept2'])+2])

plt.show()

代码说明 #

• 交换hue和style参数后，每个Shorthand条目对应唯一颜色，每个Subscale对应唯一形状，解决原图表颜色占比失衡的问题
• 通过style_dict提取Subscale对应的标记形状，再将其分配给对应的Shorthand图例条目，实现形状与颜色的关联
• 保留原有的配色方案（色弱友好的colorblind调色板）和布局配置，确保图表可读性与专业性

内容的提问来源于stack exchange，提问作者Aristle Monteiro