- Navigate to the Login Page of your platform instance (for example https://app.external-public.agri-gaia.edvsz.hs-osnabrueck.de/)
- Click “Register”
- Enter your personal Information
- Make sure, your Username only consists of lower case letters and no special signs
- Click Register
- Upload a dataset containing the provided images (location of the images):
- Make sure to select the correct dataset type when uploading: “AgriImageDataResource”.
- Select all provided images in the file upload (do NOT upload the annotation file here).
- Add the annotation file separately in the designated field during the upload process.
- Add suitable keywords under which the dataset should be searchable (e.g., Potatoes or Kartoffeln). The language does not matter here. In the background, Agrovoc is used to map the keyword to a higher-level concept. If your term cannot be found in Agrovoc, an appropriate message will appear. (Note: The suggested autocomplete options are only available in English.)
- On the second page of the upload form, add some corresponding metadata to the dataset, such as the number of images or image format. These metadata will not be used further in this tutorial but are mainly intended for data exchange between platform instances. Therefore, you can also enter placeholders, as long as the data type (integer, string, etc.) matches.
- Ensure that your dataset can be found using the search functionality:
- Search for the keywords you added to the dataset.
- Use the search bar to navigate through the different higher-level and subcategories displayed for your keyword. You will notice that your dataset is also found if a higher-level category of your keyword is searched. Additionally, search is possible in multiple languages.
- Add labels to the dataset:
- Open your dataset in the platform’s internal CVAT annotation tool (via the action button on the right side of your dataset row).
- A job will be displayed there. Select it and add the missing bounding boxes around the objects (for testing purposes, mark the 3 missing potatoes as good / no_defect and one as bad / defect).
- Save the dataset including annotations.
Optional:
- Check where your dataset has been stored in the internal object storage (menu on the left → Object Store).
- Check which metadata has been created for your dataset (menu on the left → Triple Store). You can run the following query in the dataset
ds. Replace the ID in the query with the ID of your dataset. You can easily find this ID by looking at the folder name in your MinIO bucket.
PREFIX rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns#
PREFIX rdfs: http://www.w3.org/2000/01/rdf-schema#
SELECT ?sub ?pred ?obj WHERE {
?sub ?pred ?obj .
Filter (?sub = <https://minio:9000/[Nutzername]/datasets/[ID des Datensatzes]#_Dataset>) } LIMIT 100