"Just draw some data and get on with your day."
This small Python library contains Jupyter widgets that allow you to draw a dataset in a Jupyter notebook. This should be very useful when teaching machine learning algorithms.
The project uses anywidget under the hood so our tools should work in Jupyter, VSCode and Colab. That also means that you get a proper widget that can interact with ipywidgets natively. Here is an example where updating a drawing triggers a new scikit-learn model to train (code).
You can really get creative with this in a notebook, so feel free to give it a spin!
Installation occurs via pip.
python -m pip install drawdata
To read the data, polars is useful, but this library also suppots pandas:
python -m pip install pandas polars
You can load the scatter widget to start drawing immediately.
from drawdata import ScatterWidget
widget = ScatterWidget()
widgetIf you want to use the dataset that you've just drawn you can do so via:
# Get the drawn data as a list of dictionaries
widget.data
# Get the drawn data as a dataframe
widget.data_as_pandas
widget.data_as_polarsIf you're eager to do scikit-learn stuff with your drawn data you may appreciate this property instead:
X, y = widget.data_as_X_y
The assumption for this property is that if you've used multiple colors that you're interested in doing classification and if you've only drawn one color you're interested in regression. In the case of regression y will refer to the y-axis.
This project was originally part of my work over at calmcode labs but my employer probabl has been very supportive and has allowed me to work on this project during my working hours. This was super cool and I wanted to make sure I recognise them for it.
The original implementation of our widget would use an iframe to load a site in order
to be able to draw from a Jupyter notebook. This works, but requires more manual effort, only works with pandas via the clipboard feature and needs an internet connection. Here's what that widget looks like:
It will be kept around, but the way forward for this library is to build on top of anywidget.
When you run this from jupyter, you should load in an iframe.
from drawdata import draw_scatter
draw_scatter()Once you're done drawing you can copy the data to the clipboard. After this you can use pandas to read the clipboard to get your drawn data into a dataframe.
import pandas as pd
pd.read_clipboard(sep=",")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent