Blue Noise Plots

Python implementation of Blue Noise Plots, a novel replacement for jitter plots, published at Eurographics 2021.

C. van Onzenoodt, G. Singh, T. Ropinski, and T. Ritschel

Paper Pre-Print

Project Page

Prerequisites

This implementation uses tensorflow to enable hardware acceleration on supported platforms (Linux and Windows). Therefore, Blue Noise Plots are currently limited by prerequisites of tensorflow:

• Python 3.5–3.8
• Ubuntu 16.04 or later
• Windows 7 or later
• macOS 10.12.6 (Sierra) or later (no GPU support)

Install dependencies

cd BlueNoisePlots
pip install -r requirements.txt

Usage

import pandas as pd
from blue_noise_plot import blue_noise

# Load data
mpg_filename = 'csv_data/auto-mpg.data'
column_names = ['mpg', 'cylinders', 'displacement', 'horsepower', 'weight',
                'acceleration', 'year', 'origin', 'name']
mpg = pd.read_csv(mpg_filename, delim_whitespace=True, names=column_names)
subset = mpg[mpg['cylinders'].isin([4, 6, 8])]


# Blue noise plot of `subset`, using our automatic width computation
points = blue_noise(x='mpg', data=subset, orient='h')
print('Num Classes: ', len(points))   # Num Classes:  1
print('Num Points: ', len(points[0])) # Num Points:  391

points = blue_noise(x='mpg', hue='cylinders', data=subset, orient='h')
print('Num Classes: ', len(points))   # Num Classes:  3
print('Num Points: ', len(points[0])) # Num Points:  204
print('Num Points: ', len(points[1])) # Num Points:  103
print('Num Points: ', len(points[2])) # Num Points:  84

points = blue_noise(x='mpg', hue='cylinders', data=subset, orient='h')

# Blue noise plot of `subset`, using predefined width.
points = blue_noise(x='mpg', hue='cylinders', data=subset, orient='h', plot_width=0.3)

# Render png of the distribution
blue_noise(x='mpg', hue='cylinders', data=subset, orient='h', size=20, 
           filename='mpg-blue_noise_plot.png')

Advanced Usage

# Centralized Blue Noise Plot (see Examples below)
blue_noise(x='mpg', hue='cylinders', data=subset, centralized=True,
           orient='h', size=20, filename='mpg-blue_noise_plot.png')
           
# Dodged Blue Noise Plot (see Examples below)
blue_noise(x='mpg', hue='cylinders', data=subset, dodge=True,
           orient='h', size=20, filename='mpg-blue_noise_plot.png')

Examples

To generate the examples, run the following:

cd examples
pip install -r requirements.txt
./example_plots.py