Bokeh and HoloViews for labeling Machine Learning data in Python
presentationposted on 30.01.2020 by James A. Bednar
Academic presentations can be uploaded in their original slide format. Presentations are usually represented as slide decks. Videos of presentations can be uploaded as media.
A variety of special-purpose tools like https://github.com/tzutalin/labelImg are available for annotating/tagging/labeling data for machine-learning pipelines.
If those work well for your project, great! But what if you need something more customized to your particular workflow? Do you have to build a new tool from scratch? Not if you use Python, where you can now easily build custom apps based on Bokeh.org's drawing tools to collect user input. This talk goes through example Python code in a Jupyter notebook available from https://examples.pyviz.org/ml_annotators/ml_annotators.html . These examples show how to build custom labeling apps using the high-level holoviews.org/geoviews.org interfaces that can make full-featured custom tools in just a few lines of Python code, fitting naturally into your Python-based data-processing pipelines.
This presentation was given at the Earth Science Information Partners (ESIP) Winter Meeting held in Bethesda, MD in January 2020.