Imagine editing a picture layer by layer, not pixel by pixel, nor by splashing paint.

We can apply this idea to datasets.

Draw shapes and apply labels with

hover

• A view from above: plot your dataset in 2D with tooltips

  • See the whole picture while able to inspect every detail.
  • Know how much data has been labeled and where they are.

• Balance between speed and precision

  • Come up with labels as you inspect the data, no need to pre-define them.
  • Identify & label groups of data of any shape or scale.
  • Leverage additional filter conditions such as regex or soft label score.

Check out @phurwicz/hover-binder for a list of demo apps.

:flags: Project News

• Apr 30, 2021 0.5.0 is now available. Check out the changelog for details :partying_face:. Some tl-dr for the impatient:
  • you can now filter selected data with search criteria, or soft label scores, or both!
  • active learning now includes an interpolation between input and output manifolds, helping you explore decision boundaries and their formation.

:rocket: Quickstart

• Example script with walkthrough
• Example on Binder

:package: Install

Python: 3.6+

OS: tested on Mac & Linux

To get the latest release version:

pip install hover

Feel free to open an issue if you would like

conda

conda-forge

:book: Resources

• Binder repo
• Changelog
• Documentation
• Tutorials

:bell: Remarks

Shoutouts

• Thanks to

  Bokeh

  because

  hover

  would not exist without linked plots and callbacks, or be nearly as good without embeddable server apps.
• Thanks to Philip Vollet for sharing

  hover

  with the community even when it was really green.

Contributing

• All feedbacks are welcome :hugs: Especially what you find frustrating and want fixed!

Developer

• ./requirements-dev.txt

  lists recommended packages for development.
• You are encouraged, but not required, to use pre-commit hooks.