Annotating pictures (i.e. images, imagery, or still imagery) is a core part of the platform. Our platform's Annotation Interface has all the tools you'll need to quickly create high-quality training data you can use over and over again on your machine learning journey.
This article assumes you've already set up a Project and are ready to start annotating. If that's not true, head back to our top-level article about that.
Get started: The Basics
Here is a snapshot of our annotation interface when annotating an image:
On the left-hand sidebar you will see (in order from top to bottom):
Submit task button
Skip or Reject task
If a task is difficult or you have a question on it for your team, you can skip the task and come back to it later. The skipped tasks will be in your task assignments area on the annotate tab in the project or your annotation tasks tab in the platform.
If there is any reason you might want to reject the task and remove it from your project (but keep it in the original dataset) you can click to reject the task.
Add/Edit Button
If you want to add a label, make sure you are in add mode or click the "A" key. If you want to edit a label, then you can click into edit mode or click the "E" key.
Instructions
This will pull up the instructions you have attached to the project for a refresher on what each category is or other key notes about annotation.
Validation Issues
If there are any issues in the labels you have added, you will see a red square here with the number of issues found.
For EX. When you are labeling an object for a segmentation model, if you have lines that are intersecting for one object that will bring up a validation issue for you to correct.
Below is an example of what the error may look like with intersecting lines:
Here is what the correct label should look like to avoid the issue:
Tool Box
This is an area where you can find other helpful tools to use in annotation.
The toolbox has a measurement tool to measure distances in pixels, an adjacent capture tool to see other images that would be next to your task in geospatial data that has been tiled, an angle tool for helping create perfect lines/angles, a color by capture tool to select all labels with a specific color, and a cut tool to cut a label in half or cut out a piece of the label you created.
For geospatial data ONLY: you will see google maps and mapbox which will bring you to the exact location on that data on the map to get a better idea of what you are looking at.
Lastly you will see the shortcuts keyboard at the very bottom. This will bring up all the hot keys for quicker labeling.
In the center you will see your image task
This is the area you will be adding labels too!
On the right-hand side of the screen, you will see:
Zoom in/out buttons
All of the classes you can label in the project
once you start labeling each time you add a label for a class you will see the number adjust next to that specific class.
you can also click the eye to hide the labels you have already done for that class
there is also a lock key to keep those annotations intact while you are editing other labels.
Loom Videos for annotation of each model architecture type: