Let’s start by creating a new project. Go to the ‘Projects’ tab at the top and click ‘Create project’. Click "Next".
Project name and description
Give your project an appropriate name and description.
You'll see the dataset setup page below. A dataset is a collection of videos and images that you want to use in your project to annotate and/or train your models on.
Here we can select an existing dataset or create a new dataset that you can then import into the project. When creating a new dataset, name the dataset and upload the data you want it to include. When it is done uploading, it will appear in the available datasets to include in the project.
To upload images, click the 'Upload images' button to toggle the 'Create image group' modal. An image group is a collection of compressed images. This allows for more efficient annotation of semantically similar images, as you can use automation features such as object tracking algorithms to aid you in the annotation process.
Once you have uploaded your selected images, click on 'Create' to create the image group. The resultant image group is automatically added to the dataset, and you can still label each image individually.
Label structure (ontology)
Now that we have imported data in the project, we can set up our label structure (ontology).
Adding objects and classifications
Cord offers support for bounding boxes and polygonal objects as well as frame level classifications, all with arbitrarily deep nested attributes. To get started, add an item under either ‘Objects’ or ‘Classifications’. For an item under ‘Objects’ select ‘Polygon’ or ‘Box’ under the dropdown on the right and configure the object boundary colour with the circle icon on the left.
Adding nested attributes
To add a nested attribute, click on the right-hand blue ‘Configure’ button. This will take you to attribute configuration. Write the answer attribute name under ‘Instructions’
Here you can add attribute options in ‘Text’, ‘Radio’ or ‘Checklist’ forms for the object. Add as many ‘Radio’ or ‘Checklist’ options as you need. Additionally, add further nested attributes under each option recursively by clicking the blue ‘Nested options’ button on the right of the option. This allows you to create a nested tree of attributes that can describe your objects with as much complexity as you need.
Similarly, add nested frame level classifications by adding a ‘Classifications’ item. Input the classification title at the top and add ‘Text’ , ‘Radio’ , or ‘Checklist’ and follow the same recursive logic as within the ‘Objects’ attributes.
Once you have configured your ontology to your satisfaction, preview the tree structure by toggling the ‘Display json’ toggle on the upper right.
After reviewing the structure, go ahead and click ‘Create’ and start working on your first project!