/agdss

decision support system for robotic sampling in precision agriculture

Primary LanguageHTMLApache License 2.0Apache-2.0

AgDSS

Django based Agricultural Decision Support System (AgDSS) for robotic sampling in precision agriculture. Please note that this documentation is incomplete and may be out of date.

Setup

  1. git clone https://github.com/Trefo/agdss
  2. cd agdss

The folder annotable/small-tomatoes has a set of example images. Accordingly, the agdss/settings/common.py file assumes small-tomatoes as category. You need to specify your own folder and image category you want to label, as well as the name of folder where converted image label masks will be stored (default is 'labels/' in commons.py). Once, satisfied run docker-compose!

  1. docker-compose up

First-time additional steps needed in agdss web container named agdss_web

On a different terminal.

  1. docker exec -it agdss_web_1 bash
  2. python manage.py migrate --settings=agdss.settings.common to create the tables in the database.
  3. python manage.py createsuperuser --settings=agdss.settings.common and enter the requested information.
  4. python manage.py collectstatic --settings=agdss.settings.common
  5. cp /app/addImages.sh /app/static-root/small-tomatoes
  6. cd /app/static-root/small-tomatoes/
  7. bash addImages.sh "[\"small-tomatoes\"]"

On using the annotation web app

The web app can then be accessed at http://172.10.0.3:8000.

Admin page is at http://172.10.0.3:8000/admin Use the credential you created for the superuser in item 3. for first-time additional steps.

For a list of available pages, see the file agdss/urls.py

Viewing labels

Labels are stored in the folder specified in the settings file, default being 'labels', which should be in the static-root folder. Labels can also be inspected through the web app using the admin image labels page.

UI Controls

You can resize the browser using standard browser controls (Ctrl+ and Ctrl-). Additionally, the app has sliders to control brightness, contrast, and hue. Adjust these to make the target objects clearer.

Labels can be moved by dragging, and deleted by double-clicking.

Submit when finished labeling an image.

Choose 'No labels in image' when there are no labels so that that the data is recorded correctly on the DB.