A simple Flask webapp for categorical pixel labeling of histological slides. Written in Python (Flask), JavaScript.

The latest version of Python 3 is recommended.

This application was built and tested with PostgreSQL 13.2. After installing PostgreSQL package, a new database needs to be created. For more information on how to setup a new PostgreSQL database visit https://wiki.archlinux.org/title/PostgreSQL.

Using virtual environments for managing dependencies for a project is always recommended.

$ python -m venv path/to/your/virtualenvironment

$ source path/to/your/virtualenvironment/bin/activate

$ pip install -r requirements.txt

In .flaskenv file, a variable DATABASE_URL needs to be correctly declared depending on the name of your newly created database.

Each time a change is made to the database schema (app/models.py), a migration script is added to the repository with the details of the change. To apply the migrations to a database, these migration scripts are executed in the sequence they were created.

$ flask db init # only run when first creating a migration repository
$ flask db migrate 
$ flask db upgrade

From the folder you cloned the repository into run flask run.

• This application is based on the idea that a user creates a new project by adding images to be annotated by themselves or other other users who are added to a project by the creator of it.
• Project creator can add other users by specifying their username.
• After a new user is added to a project, it will show up in their list of projects and they can start labeling.

• When logged in, user is redirected to their home page where they can see projects they created or projects they were added to by choosing the corresponding tab (Not implemented)

• In the current version, a brush tool with adjustable radius or one of provided superpixel algorithms with some adjustable parameters are supported.

  Superpixels

  • 4 algorithms are currently implemented
  • For SLIC, a number of segments for a picture to be divided into can be set and compactness, ie. higher value makes superpixel shapes more cubic.
  • After making changes to either one of these parameters or choosing a different superpixel algorithm, Calculate segments applies it.

• Right clicking with either brush tool or a superpixel tool, deletes selected area of the mask.

• Clear button clears whole image mask.

• Image order cannot be changed. Only after a loaded image is finished labeling by clicking on the Save groundtruth and next image button, next one will be loaded.

• If a user cannot finish an image in a session, they can choose to Save checkpoint and logout, which will load current progress when a user logs back in and continues labeling.

Project creator can see the progress of all other participants by clicking on the + button and can download all masks by clicking the download masks icon.

Database tables are defined in models.py and SQLAlchemy is used for object-relational mapping between classes and PostgreSQL database entities.

Endpoints of the Flask app are defined in routes.py. The Flask-WTF extension uses Python classes to represent web forms. A form class simply defines the fields of the form as class variables. These are stored in forms.py

Main logic for drawing on HTML5 Canvas element which is overlaying the photo, is stored in app/static/js/drawing.js. Sending and recieving data via POST method is implemented in posting.js.

• Implement save checkpoint
• Superpixel segments do not get received by frontend
• New user registration should be approved by IRB staff
• Database needs refactoring because of some cascading issues
• Different layouts on home screen for projects created by user and shared with them
• Add some kind of waiting indicator when an image is finished
• When user labels the last image of a project, redirect to home
• When creating a new project, modal box closes even if there are errors in the name of the project
• Fix background image on login screen