Explore data with one line of code

A simple interactive plotting library

click+drag to
select data

Dataset:

wiki-art

mnist

cifar10

librispeech

Plot:

Reproduce with:

import pandas as pd
import clusterfun as clt

df = pd.read_csv("https://raw.githubusercontent.com/gietema/clusterfun-data/main/mnist.csv")
clt.scatter(df, x="x", y="y", media="img_path", color="label")

import pandas as pd
import clusterfun as clt

df = pd.read_csv("https://raw.githubusercontent.com/gietema/clusterfun-data/main/mnist.csv")
clt.confusion_matrix(df, y_true="label", y_pred="pred", media="img_path")

import pandas as pd
import clusterfun as clt

df = pd.read_csv("https://raw.githubusercontent.com/gietema/clusterfun-data/main/mnist.csv")
clt.violin(df, y="mean_value", color="label", media="img_path")

That's it!

Why clusterfun?

Lightweight
As opposed to big data analysis platforms, clusterfun allows you to review data within a minute. All you need is a pandas dataframe with an image/audio location per row. Data can be hosted locally or on AWS S3. No sign up, integration, or data transfer is needed.
Flexible
Clusterfun is a plotting library, which means that you determine what the x- and y-axis represent. This gives you the freedom to plot what you want. The additional benefit of a plot is that it allows you to see how close a data point is to another data point, which is more difficult in an image grid visualisation.
Interactive
As opposed to static plots, clusterfun allows you to hover, select, and zoom in on data slices.

Documentation

Data exploration

Labelling

Model understanding

Error analysis

Explore data with one line of code

A simple interactive plotting library

Clusterfun

By Gietema