The challenges of developing research tools in active and responsive open source communities

This poster presents work by Professorial Assistance, Ty Buckly and Hattie Pimentel and their work on building an image annotation system inside of Jupyter.

Abstract

Scientists often gather observations of the world using images and need efficient image annotating tools to extract data from these images. Project Insight seeks to make this process easier and quicker. The project consists of two parts: first, an image annotation package, the Graphical User Interface (GUI); and second, behind-the-scenes machine learning. When Project Insight is completed, data from the GUI will be sent to the machine learning component, which will attempt to anticipate the user’s next choices. The machine-learning component will search through algorithm space for the most suitable algorithm.

Jupyter Notebooks are an interface used by researchers to combines code, pictures, notes and other multimedia. Using Jupyter Notebooks, several image annotation tools were created. Currently, the image annotation tools are being moved to a future replacement of Jupyter Notebooks: JupyterLab. Citing security concerns, the JupyterLab team blocked key JavaScript execution in the beta version of JupyterLab. This means that many programs that worked in the original Jupyter Notebooks, including most of the Project Insight programs, are not functional in JupyterLab.

This presentation will cover multiple attempts to retaining functionality in JupyterLab and discuss why the JupyterLab’s security block hinders functionality and contrasts the accessible theme of JupyterLab.

Link to the poster